Person Tracking with OpenVINO™

This Jupyter notebook can be launched on-line, opening an interactive environment in a browser window. You can also make a local installation. Choose one of the following options:

This notebook demonstrates live person tracking with OpenVINO: it reads frames from an input video sequence, detects people in the frames, uniquely identifies each one of them and tracks all of them until they leave the frame. We will use the Deep SORT algorithm to perform object tracking, an extension to SORT (Simple Online and Realtime Tracking).

Detection vs Tracking

• In object detection, we detect an object in a frame, put a bounding box or a mask around it, and classify the object. Note that, the job of the detector ends here. It processes each frame independently and identifies numerous objects in that particular frame.

• An object tracker on the other hand needs to track a particular object across the entire video. If the detector detects three cars in the frame, the object tracker has to identify the three separate detections and needs to track it across the subsequent frames (with the help of a unique ID).

Deep SORT

Deep SORT can be defined as the tracking algorithm which tracks objects not only based on the velocity and motion of the object but also the appearance of the object. It is made of three key components which are as follows:

1. Detection

   This is the first step in the tracking module. In this step, a deep learning model will be used to detect the objects in the frame that are to be tracked. These detections are then passed on to the next step.

2. Prediction

   In this step, we use Kalman filter [1] framework to predict a target bounding box of each tracking object in the next frame. There are two states of prediction output: confirmed and unconfirmed. A new track comes with a state of unconfirmed by default, and it can be turned into confirmed when a certain number of consecutive detections are matched with this new track. Meanwhile, if a matched track is missed over a specific time, it will be deleted as well.

3. Data association and update

   Now, we have to match the target bounding box with the detected bounding box, and update track identities. A conventional way to solve the association between the predicted Kalman states and newly arrived measurements is to build an assignment problem with the Hungarian algorithm [2]. In this problem formulation, we integrate motion and appearance information through a combination of two appropriate metrics. The cost used for the first matching step is set as a combination of the Mahalanobis and the cosine distances. The Mahalanobis distance is used to incorporate motion information and the cosine distance is used to calculate similarity between two objects. Cosine distance is a metric that helps the tracker recover identities in case of long-term occlusion and motion estimation also fails. For this purposes, a reidentification model will be implemented to produce a vector in high-dimensional space that represents the appearance of the object. Using these simple things can make the tracker even more powerful and accurate.

   In the second matching stage, we will run intersection over union(IOU) association as proposed in the original SORT algorithm [3] on the set of unconfirmed and unmatched tracks from the previous step. If the IOU of detection and target is less than a certain threshold value called IOUmin then that assignment is rejected. This helps to account for sudden appearance changes, for example, due to partial occlusion with static scene geometry, and to increase robustness against erroneous.

   When detection result is associated with a target, the detected bounding box is used to update the target state.

---

[1] R. Kalman, “A New Approach to Linear Filtering and Prediction Problems”, Journal of Basic Engineering, vol. 82, no. Series D, pp. 35-45, 1960.

[2] H. W. Kuhn, “The Hungarian method for the assignment problem”, Naval Research Logistics Quarterly, vol. 2, pp. 83-97, 1955.

[3] A. Bewley, G. Zongyuan, F. Ramos, and B. Upcroft, “Simple online and realtime tracking,” in ICIP, 2016, pp. 3464–3468.

Table of contents:

• Imports

• Download the Model

• Load model

  • Select inference device

• Data Processing

• Test person reidentification model

  • Visualize data

  • Compare two persons

• Main Processing Function

• Run

  • Initialize tracker

  • Run Live Person Tracking

%pip install -q "openvino-dev>=2023.1.0"
%pip install -q opencv-python matplotlib requests scipy

DEPRECATION: pytorch-lightning 1.6.5 has a non-standard dependency specifier torch>=1.8.*. pip 24.1 will enforce this behaviour change. A possible replacement is to upgrade to a newer version of pytorch-lightning or contact the author to suggest that they release a version with a conforming dependency specifiers. Discussion can be found at https://github.com/pypa/pip/issues/12063

Note: you may need to restart the kernel to use updated packages.

DEPRECATION: pytorch-lightning 1.6.5 has a non-standard dependency specifier torch>=1.8.*. pip 24.1 will enforce this behaviour change. A possible replacement is to upgrade to a newer version of pytorch-lightning or contact the author to suggest that they release a version with a conforming dependency specifiers. Discussion can be found at https://github.com/pypa/pip/issues/12063

Note: you may need to restart the kernel to use updated packages.

Imports

import collections
from pathlib import Path
import sys
import time

import numpy as np
import cv2
from IPython import display
import matplotlib.pyplot as plt
import openvino as ov

# Import local modules

utils_file_path = Path('../utils/notebook_utils.py')
notebook_directory_path = Path('.')

if not utils_file_path.exists():
    !git clone --depth 1 https://github.com/igor-davidyuk/openvino_notebooks.git -b moving_data_to_cloud openvino_notebooks
    utils_file_path = Path('./openvino_notebooks/notebooks/utils/notebook_utils.py')
    notebook_directory_path = Path('./openvino_notebooks/notebooks/407-person-tracking-webcam/')

sys.path.append(str(utils_file_path.parent))
sys.path.append(str(notebook_directory_path))

import notebook_utils as utils
from deepsort_utils.tracker import Tracker
from deepsort_utils.nn_matching import NearestNeighborDistanceMetric
from deepsort_utils.detection import Detection, compute_color_for_labels, xywh_to_xyxy, xywh_to_tlwh, tlwh_to_xyxy

Download the Model

We will use pre-trained models from OpenVINO’s Open Model Zoo to start the test.

Use omz_downloader, which is a command-line tool from the openvino-dev package. It automatically creates a directory structure and downloads the selected model. This step is skipped if the model is already downloaded. The selected model comes from the public directory, which means it must be converted into OpenVINO Intermediate Representation (OpenVINO IR).

NOTE: Using a model outside the list can require different pre- and post-processing.

In this case, person detection model is deployed to detect the person in each frame of the video, and reidentification model is used to output embedding vector to match a pair of images of a person by the cosine distance.

If you want to download another model (person-detection-xxx from Object Detection Models list, person-reidentification-retail-xxx from Reidentification Models list), replace the name of the model in the code below.

# A directory where the model will be downloaded.
base_model_dir = "model"
precision = "FP16"
# The name of the model from Open Model Zoo
detection_model_name = "person-detection-0202"

download_command = f"omz_downloader " \
                   f"--name {detection_model_name} " \
                   f"--precisions {precision} " \
                   f"--output_dir {base_model_dir} " \
                   f"--cache_dir {base_model_dir}"
! $download_command

detection_model_path = f"model/intel/{detection_model_name}/{precision}/{detection_model_name}.xml"


reidentification_model_name = "person-reidentification-retail-0287"

download_command = f"omz_downloader " \
                   f"--name {reidentification_model_name} " \
                   f"--precisions {precision} " \
                   f"--output_dir {base_model_dir} " \
                   f"--cache_dir {base_model_dir}"
! $download_command

reidentification_model_path = f"model/intel/{reidentification_model_name}/{precision}/{reidentification_model_name}.xml"

################|| Downloading person-detection-0202 ||################

========== Downloading model/intel/person-detection-0202/FP16/person-detection-0202.xml

... 12%, 32 KB, 1342 KB/s, 0 seconds passed

... 25%, 64 KB, 1352 KB/s, 0 seconds passed

… 38%, 96 KB, 1176 KB/s, 0 seconds passed

... 51%, 128 KB, 1385 KB/s, 0 seconds passed

… 64%, 160 KB, 1372 KB/s, 0 seconds passed … 77%, 192 KB, 1513 KB/s, 0 seconds passed … 89%, 224 KB, 1759 KB/s, 0 seconds passed … 100%, 248 KB, 1949 KB/s, 0 seconds passed

========== Downloading model/intel/person-detection-0202/FP16/person-detection-0202.bin

... 0%, 32 KB, 1296 KB/s, 0 seconds passed

… 1%, 64 KB, 1293 KB/s, 0 seconds passed … 2%, 96 KB, 1907 KB/s, 0 seconds passed

... 3%, 128 KB, 1703 KB/s, 0 seconds passed

… 4%, 160 KB, 2107 KB/s, 0 seconds passed … 5%, 192 KB, 2285 KB/s, 0 seconds passed … 6%, 224 KB, 2641 KB/s, 0 seconds passed

... 7%, 256 KB, 2336 KB/s, 0 seconds passed

… 8%, 288 KB, 2606 KB/s, 0 seconds passed … 9%, 320 KB, 2703 KB/s, 0 seconds passed … 9%, 352 KB, 2953 KB/s, 0 seconds passed … 10%, 384 KB, 2663 KB/s, 0 seconds passed … 11%, 416 KB, 2873 KB/s, 0 seconds passed … 12%, 448 KB, 2930 KB/s, 0 seconds passed … 13%, 480 KB, 3123 KB/s, 0 seconds passed

... 14%, 512 KB, 2867 KB/s, 0 seconds passed

… 15%, 544 KB, 2911 KB/s, 0 seconds passed … 16%, 576 KB, 3076 KB/s, 0 seconds passed … 17%, 608 KB, 3233 KB/s, 0 seconds passed

... 18%, 640 KB, 3007 KB/s, 0 seconds passed

… 18%, 672 KB, 3146 KB/s, 0 seconds passed … 19%, 704 KB, 3173 KB/s, 0 seconds passed … 20%, 736 KB, 3306 KB/s, 0 seconds passed … 21%, 768 KB, 3105 KB/s, 0 seconds passed … 22%, 800 KB, 3224 KB/s, 0 seconds passed … 23%, 832 KB, 3245 KB/s, 0 seconds passed … 24%, 864 KB, 3362 KB/s, 0 seconds passed

... 25%, 896 KB, 3181 KB/s, 0 seconds passed

… 26%, 928 KB, 3286 KB/s, 0 seconds passed … 27%, 960 KB, 3303 KB/s, 0 seconds passed … 27%, 992 KB, 3404 KB/s, 0 seconds passed

... 28%, 1024 KB, 3237 KB/s, 0 seconds passed

… 29%, 1056 KB, 3331 KB/s, 0 seconds passed … 30%, 1088 KB, 3345 KB/s, 0 seconds passed … 31%, 1120 KB, 3436 KB/s, 0 seconds passed … 32%, 1152 KB, 3287 KB/s, 0 seconds passed … 33%, 1184 KB, 3371 KB/s, 0 seconds passed … 34%, 1216 KB, 3384 KB/s, 0 seconds passed … 35%, 1248 KB, 3464 KB/s, 0 seconds passed

... 36%, 1280 KB, 3323 KB/s, 0 seconds passed

… 36%, 1312 KB, 3335 KB/s, 0 seconds passed … 37%, 1344 KB, 3410 KB/s, 0 seconds passed … 38%, 1376 KB, 3484 KB/s, 0 seconds passed

... 39%, 1408 KB, 3355 KB/s, 0 seconds passed

… 40%, 1440 KB, 3428 KB/s, 0 seconds passed … 41%, 1472 KB, 3436 KB/s, 0 seconds passed … 42%, 1504 KB, 3505 KB/s, 0 seconds passed … 43%, 1536 KB, 3383 KB/s, 0 seconds passed … 44%, 1568 KB, 3442 KB/s, 0 seconds passed … 45%, 1600 KB, 3457 KB/s, 0 seconds passed … 45%, 1632 KB, 3518 KB/s, 0 seconds passed

... 46%, 1664 KB, 3404 KB/s, 0 seconds passed

… 47%, 1696 KB, 3462 KB/s, 0 seconds passed … 48%, 1728 KB, 3475 KB/s, 0 seconds passed … 49%, 1760 KB, 3532 KB/s, 0 seconds passed

... 50%, 1792 KB, 3427 KB/s, 0 seconds passed

… 51%, 1824 KB, 3480 KB/s, 0 seconds passed … 52%, 1856 KB, 3491 KB/s, 0 seconds passed … 53%, 1888 KB, 3545 KB/s, 0 seconds passed … 54%, 1920 KB, 3444 KB/s, 0 seconds passed … 54%, 1952 KB, 3493 KB/s, 0 seconds passed … 55%, 1984 KB, 3505 KB/s, 0 seconds passed … 56%, 2016 KB, 3555 KB/s, 0 seconds passed

... 57%, 2048 KB, 3457 KB/s, 0 seconds passed

… 58%, 2080 KB, 3466 KB/s, 0 seconds passed … 59%, 2112 KB, 3515 KB/s, 0 seconds passed … 60%, 2144 KB, 3562 KB/s, 0 seconds passed

... 61%, 2176 KB, 3471 KB/s, 0 seconds passed

… 62%, 2208 KB, 3480 KB/s, 0 seconds passed … 63%, 2240 KB, 3525 KB/s, 0 seconds passed … 64%, 2272 KB, 3570 KB/s, 0 seconds passed … 64%, 2304 KB, 3484 KB/s, 0 seconds passed … 65%, 2336 KB, 3492 KB/s, 0 seconds passed … 66%, 2368 KB, 3533 KB/s, 0 seconds passed … 67%, 2400 KB, 3577 KB/s, 0 seconds passed

... 68%, 2432 KB, 3494 KB/s, 0 seconds passed

… 69%, 2464 KB, 3501 KB/s, 0 seconds passed … 70%, 2496 KB, 3543 KB/s, 0 seconds passed … 71%, 2528 KB, 3585 KB/s, 0 seconds passed

... 72%, 2560 KB, 3503 KB/s, 0 seconds passed

… 73%, 2592 KB, 3509 KB/s, 0 seconds passed … 73%, 2624 KB, 3549 KB/s, 0 seconds passed … 74%, 2656 KB, 3589 KB/s, 0 seconds passed … 75%, 2688 KB, 3512 KB/s, 0 seconds passed … 76%, 2720 KB, 3518 KB/s, 0 seconds passed … 77%, 2752 KB, 3557 KB/s, 0 seconds passed

... 78%, 2784 KB, 3596 KB/s, 0 seconds passed

… 79%, 2816 KB, 3525 KB/s, 0 seconds passed … 80%, 2848 KB, 3531 KB/s, 0 seconds passed … 81%, 2880 KB, 3564 KB/s, 0 seconds passed … 82%, 2912 KB, 3601 KB/s, 0 seconds passed

... 82%, 2944 KB, 3529 KB/s, 0 seconds passed

… 83%, 2976 KB, 3534 KB/s, 0 seconds passed … 84%, 3008 KB, 3569 KB/s, 0 seconds passed … 85%, 3040 KB, 3508 KB/s, 0 seconds passed … 86%, 3072 KB, 3536 KB/s, 0 seconds passed

... 87%, 3104 KB, 3539 KB/s, 0 seconds passed

… 88%, 3136 KB, 3573 KB/s, 0 seconds passed … 89%, 3168 KB, 3608 KB/s, 0 seconds passed … 90%, 3200 KB, 3544 KB/s, 0 seconds passed … 91%, 3232 KB, 3549 KB/s, 0 seconds passed … 91%, 3264 KB, 3578 KB/s, 0 seconds passed … 92%, 3296 KB, 3612 KB/s, 0 seconds passed

... 93%, 3328 KB, 3553 KB/s, 0 seconds passed

… 94%, 3360 KB, 3555 KB/s, 0 seconds passed … 95%, 3392 KB, 3583 KB/s, 0 seconds passed … 96%, 3424 KB, 3529 KB/s, 0 seconds passed … 97%, 3456 KB, 3554 KB/s, 0 seconds passed

... 98%, 3488 KB, 3558 KB/s, 0 seconds passed

… 99%, 3520 KB, 3588 KB/s, 0 seconds passed … 100%, 3549 KB, 3616 KB/s, 0 seconds passed

################|| Downloading person-reidentification-retail-0287 ||################

========== Downloading model/intel/person-reidentification-retail-0287/person-reidentification-retail-0267.onnx

... 0%, 32 KB, 1173 KB/s, 0 seconds passed

... 1%, 64 KB, 1062 KB/s, 0 seconds passed

… 2%, 96 KB, 1556 KB/s, 0 seconds passed … 3%, 128 KB, 1333 KB/s, 0 seconds passed … 4%, 160 KB, 1577 KB/s, 0 seconds passed

... 5%, 192 KB, 1482 KB/s, 0 seconds passed

… 6%, 224 KB, 1719 KB/s, 0 seconds passed … 7%, 256 KB, 1959 KB/s, 0 seconds passed … 8%, 288 KB, 2199 KB/s, 0 seconds passed

... 9%, 320 KB, 1952 KB/s, 0 seconds passed

… 10%, 352 KB, 2136 KB/s, 0 seconds passed … 11%, 384 KB, 2325 KB/s, 0 seconds passed … 11%, 416 KB, 2445 KB/s, 0 seconds passed … 12%, 448 KB, 2257 KB/s, 0 seconds passed … 13%, 480 KB, 2408 KB/s, 0 seconds passed … 14%, 512 KB, 2564 KB/s, 0 seconds passed

... 15%, 544 KB, 2428 KB/s, 0 seconds passed

… 16%, 576 KB, 2473 KB/s, 0 seconds passed … 17%, 608 KB, 2602 KB/s, 0 seconds passed … 18%, 640 KB, 2734 KB/s, 0 seconds passed … 19%, 672 KB, 2867 KB/s, 0 seconds passed

... 20%, 704 KB, 2632 KB/s, 0 seconds passed

… 21%, 736 KB, 2745 KB/s, 0 seconds passed … 22%, 768 KB, 2859 KB/s, 0 seconds passed … 22%, 800 KB, 2974 KB/s, 0 seconds passed … 23%, 832 KB, 2755 KB/s, 0 seconds passed … 24%, 864 KB, 2856 KB/s, 0 seconds passed … 25%, 896 KB, 2957 KB/s, 0 seconds passed … 26%, 928 KB, 3058 KB/s, 0 seconds passed

... 27%, 960 KB, 2853 KB/s, 0 seconds passed

… 28%, 992 KB, 2943 KB/s, 0 seconds passed … 29%, 1024 KB, 3034 KB/s, 0 seconds passed … 30%, 1056 KB, 3125 KB/s, 0 seconds passed

... 31%, 1088 KB, 2933 KB/s, 0 seconds passed

… 32%, 1120 KB, 3012 KB/s, 0 seconds passed … 33%, 1152 KB, 3095 KB/s, 0 seconds passed … 33%, 1184 KB, 3142 KB/s, 0 seconds passed … 34%, 1216 KB, 2999 KB/s, 0 seconds passed … 35%, 1248 KB, 3074 KB/s, 0 seconds passed … 36%, 1280 KB, 3148 KB/s, 0 seconds passed … 37%, 1312 KB, 3190 KB/s, 0 seconds passed

... 38%, 1344 KB, 3055 KB/s, 0 seconds passed

… 39%, 1376 KB, 3124 KB/s, 0 seconds passed … 40%, 1408 KB, 3191 KB/s, 0 seconds passed … 41%, 1440 KB, 3100 KB/s, 0 seconds passed

... 42%, 1472 KB, 3102 KB/s, 0 seconds passed

… 43%, 1504 KB, 3165 KB/s, 0 seconds passed … 44%, 1536 KB, 3228 KB/s, 0 seconds passed … 44%, 1568 KB, 3143 KB/s, 0 seconds passed … 45%, 1600 KB, 3144 KB/s, 0 seconds passed … 46%, 1632 KB, 3201 KB/s, 0 seconds passed … 47%, 1664 KB, 3260 KB/s, 0 seconds passed

... 48%, 1696 KB, 3180 KB/s, 0 seconds passed

… 49%, 1728 KB, 3180 KB/s, 0 seconds passed … 50%, 1760 KB, 3234 KB/s, 0 seconds passed … 51%, 1792 KB, 3289 KB/s, 0 seconds passed … 52%, 1824 KB, 3212 KB/s, 0 seconds passed

... 53%, 1856 KB, 3210 KB/s, 0 seconds passed

… 54%, 1888 KB, 3262 KB/s, 0 seconds passed … 55%, 1920 KB, 3314 KB/s, 0 seconds passed … 55%, 1952 KB, 3194 KB/s, 0 seconds passed … 56%, 1984 KB, 3241 KB/s, 0 seconds passed … 57%, 2016 KB, 3288 KB/s, 0 seconds passed … 58%, 2048 KB, 3337 KB/s, 0 seconds passed

... 59%, 2080 KB, 3266 KB/s, 0 seconds passed

… 60%, 2112 KB, 3266 KB/s, 0 seconds passed … 61%, 2144 KB, 3310 KB/s, 0 seconds passed … 62%, 2176 KB, 3357 KB/s, 0 seconds passed … 63%, 2208 KB, 3385 KB/s, 0 seconds passed

... 64%, 2240 KB, 3289 KB/s, 0 seconds passed

… 65%, 2272 KB, 3330 KB/s, 0 seconds passed … 66%, 2304 KB, 3373 KB/s, 0 seconds passed … 66%, 2336 KB, 3267 KB/s, 0 seconds passed … 67%, 2368 KB, 3306 KB/s, 0 seconds passed … 68%, 2400 KB, 3347 KB/s, 0 seconds passed … 69%, 2432 KB, 3389 KB/s, 0 seconds passed

... 70%, 2464 KB, 3290 KB/s, 0 seconds passed

… 71%, 2496 KB, 3325 KB/s, 0 seconds passed … 72%, 2528 KB, 3364 KB/s, 0 seconds passed … 73%, 2560 KB, 3405 KB/s, 0 seconds passed

... 74%, 2592 KB, 3308 KB/s, 0 seconds passed

… 75%, 2624 KB, 3341 KB/s, 0 seconds passed … 76%, 2656 KB, 3378 KB/s, 0 seconds passed … 77%, 2688 KB, 3417 KB/s, 0 seconds passed … 77%, 2720 KB, 3323 KB/s, 0 seconds passed … 78%, 2752 KB, 3357 KB/s, 0 seconds passed … 79%, 2784 KB, 3391 KB/s, 0 seconds passed … 80%, 2816 KB, 3429 KB/s, 0 seconds passed

... 81%, 2848 KB, 3339 KB/s, 0 seconds passed

… 82%, 2880 KB, 3370 KB/s, 0 seconds passed … 83%, 2912 KB, 3404 KB/s, 0 seconds passed … 84%, 2944 KB, 3440 KB/s, 0 seconds passed

... 85%, 2976 KB, 3354 KB/s, 0 seconds passed

… 86%, 3008 KB, 3385 KB/s, 0 seconds passed … 87%, 3040 KB, 3417 KB/s, 0 seconds passed … 88%, 3072 KB, 3450 KB/s, 0 seconds passed … 88%, 3104 KB, 3366 KB/s, 0 seconds passed … 89%, 3136 KB, 3396 KB/s, 0 seconds passed

... 90%, 3168 KB, 3406 KB/s, 0 seconds passed

… 91%, 3200 KB, 3439 KB/s, 0 seconds passed … 92%, 3232 KB, 3378 KB/s, 0 seconds passed … 93%, 3264 KB, 3408 KB/s, 0 seconds passed … 94%, 3296 KB, 3417 KB/s, 0 seconds passed … 95%, 3328 KB, 3448 KB/s, 0 seconds passed

... 96%, 3360 KB, 3390 KB/s, 0 seconds passed

… 97%, 3392 KB, 3395 KB/s, 0 seconds passed … 98%, 3424 KB, 3426 KB/s, 0 seconds passed … 99%, 3456 KB, 3456 KB/s, 0 seconds passed … 100%, 3487 KB, 3432 KB/s, 1 seconds passed

========== Downloading model/intel/person-reidentification-retail-0287/FP16/person-reidentification-retail-0287.xml

... 5%, 32 KB, 1388 KB/s, 0 seconds passed

… 10%, 64 KB, 1393 KB/s, 0 seconds passed … 15%, 96 KB, 2032 KB/s, 0 seconds passed

... 21%, 128 KB, 1852 KB/s, 0 seconds passed

… 26%, 160 KB, 2275 KB/s, 0 seconds passed … 31%, 192 KB, 2375 KB/s, 0 seconds passed … 37%, 224 KB, 2717 KB/s, 0 seconds passed … 42%, 256 KB, 2459 KB/s, 0 seconds passed

... 47%, 288 KB, 2508 KB/s, 0 seconds passed

… 53%, 320 KB, 2763 KB/s, 0 seconds passed … 58%, 352 KB, 3008 KB/s, 0 seconds passed … 63%, 384 KB, 2783 KB/s, 0 seconds passed … 69%, 416 KB, 2988 KB/s, 0 seconds passed … 74%, 448 KB, 2990 KB/s, 0 seconds passed … 79%, 480 KB, 3160 KB/s, 0 seconds passed

... 85%, 512 KB, 2953 KB/s, 0 seconds passed

… 90%, 544 KB, 2958 KB/s, 0 seconds passed … 95%, 576 KB, 3112 KB/s, 0 seconds passed … 100%, 600 KB, 3237 KB/s, 0 seconds passed

========== Downloading model/intel/person-reidentification-retail-0287/FP16/person-reidentification-retail-0287.bin

... 2%, 32 KB, 1362 KB/s, 0 seconds passed

… 5%, 64 KB, 1347 KB/s, 0 seconds passed … 8%, 96 KB, 1945 KB/s, 0 seconds passed

... 11%, 128 KB, 1797 KB/s, 0 seconds passed

… 13%, 160 KB, 2222 KB/s, 0 seconds passed … 16%, 192 KB, 2330 KB/s, 0 seconds passed … 19%, 224 KB, 2685 KB/s, 0 seconds passed … 22%, 256 KB, 2414 KB/s, 0 seconds passed … 24%, 288 KB, 2704 KB/s, 0 seconds passed

... 27%, 320 KB, 2743 KB/s, 0 seconds passed

… 30%, 352 KB, 2995 KB/s, 0 seconds passed … 33%, 384 KB, 2739 KB/s, 0 seconds passed … 36%, 416 KB, 2949 KB/s, 0 seconds passed … 38%, 448 KB, 2958 KB/s, 0 seconds passed … 41%, 480 KB, 3144 KB/s, 0 seconds passed

... 44%, 512 KB, 2920 KB/s, 0 seconds passed

… 47%, 544 KB, 2936 KB/s, 0 seconds passed … 49%, 576 KB, 3093 KB/s, 0 seconds passed … 52%, 608 KB, 3259 KB/s, 0 seconds passed … 55%, 640 KB, 3060 KB/s, 0 seconds passed … 58%, 672 KB, 3195 KB/s, 0 seconds passed

... 61%, 704 KB, 3192 KB/s, 0 seconds passed

… 63%, 736 KB, 3325 KB/s, 0 seconds passed … 66%, 768 KB, 3148 KB/s, 0 seconds passed … 69%, 800 KB, 3148 KB/s, 0 seconds passed … 72%, 832 KB, 3258 KB/s, 0 seconds passed … 74%, 864 KB, 3372 KB/s, 0 seconds passed

... 77%, 896 KB, 3219 KB/s, 0 seconds passed

… 80%, 928 KB, 3322 KB/s, 0 seconds passed … 83%, 960 KB, 3318 KB/s, 0 seconds passed … 86%, 992 KB, 3417 KB/s, 0 seconds passed … 88%, 1024 KB, 3269 KB/s, 0 seconds passed … 91%, 1056 KB, 3366 KB/s, 0 seconds passed

... 94%, 1088 KB, 3360 KB/s, 0 seconds passed

… 97%, 1120 KB, 3450 KB/s, 0 seconds passed … 99%, 1152 KB, 3319 KB/s, 0 seconds passed … 100%, 1153 KB, 3319 KB/s, 0 seconds passed

Load model

Define a common class for model loading and predicting.

There are four main steps for OpenVINO model initialization, and they are required to run for only once before inference loop. 1. Initialize OpenVINO Runtime. 2. Read the network from *.bin and *.xml files (weights and architecture). 3. Compile the model for device. 4. Get input and output names of nodes.

In this case, we can put them all in a class constructor function.

To let OpenVINO automatically select the best device for inference just use AUTO. In most cases, the best device to use is GPU (better performance, but slightly longer startup time).

core = ov.Core()


class Model:
    """
    This class represents a OpenVINO model object.

    """
    def __init__(self, model_path, batchsize=1, device="AUTO"):
        """
        Initialize the model object

        Parameters
        ----------
        model_path: path of inference model
        batchsize: batch size of input data
        device: device used to run inference
        """
        self.model = core.read_model(model=model_path)
        self.input_layer = self.model.input(0)
        self.input_shape = self.input_layer.shape
        self.height = self.input_shape[2]
        self.width = self.input_shape[3]

        for layer in self.model.inputs:
            input_shape = layer.partial_shape
            input_shape[0] = batchsize
            self.model.reshape({layer: input_shape})
        self.compiled_model = core.compile_model(model=self.model, device_name=device)
        self.output_layer = self.compiled_model.output(0)

    def predict(self, input):
        """
        Run inference

        Parameters
        ----------
        input: array of input data
        """
        result = self.compiled_model(input)[self.output_layer]
        return result

select device from dropdown list for running inference using OpenVINO

import ipywidgets as widgets

device = widgets.Dropdown(
    options=core.available_devices + ["AUTO"],
    value='AUTO',
    description='Device:',
    disabled=False,
)

device

Dropdown(description='Device:', index=1, options=('CPU', 'AUTO'), value='AUTO')

detector = Model(detection_model_path, device=device.value)
# since the number of detection object is uncertain, the input batch size of reid model should be dynamic
extractor = Model(reidentification_model_path, -1, device.value)

Data Processing

Data Processing includes data preprocess and postprocess functions. - Data preprocess function is used to change the layout and shape of input data, according to requirement of the network input format. - Data postprocess function is used to extract the useful information from network’s original output and visualize it.

def preprocess(frame, height, width):
    """
    Preprocess a single image

    Parameters
    ----------
    frame: input frame
    height: height of model input data
    width: width of model input data
    """
    resized_image = cv2.resize(frame, (width, height))
    resized_image = resized_image.transpose((2, 0, 1))
    input_image = np.expand_dims(resized_image, axis=0).astype(np.float32)
    return input_image


def batch_preprocess(img_crops, height, width):
    """
    Preprocess batched images

    Parameters
    ----------
    img_crops: batched input images
    height: height of model input data
    width: width of model input data
    """
    img_batch = np.concatenate([
        preprocess(img, height, width)
        for img in img_crops
    ], axis=0)
    return img_batch


def process_results(h, w, results, thresh=0.5):
    """
    postprocess detection results

    Parameters
    ----------
    h, w: original height and width of input image
    results: raw detection network output
    thresh: threshold for low confidence filtering
    """
    # The 'results' variable is a [1, 1, N, 7] tensor.
    detections = results.reshape(-1, 7)
    boxes = []
    labels = []
    scores = []
    for i, detection in enumerate(detections):
        _, label, score, xmin, ymin, xmax, ymax = detection
        # Filter detected objects.
        if score > thresh:
            # Create a box with pixels coordinates from the box with normalized coordinates [0,1].
            boxes.append(
                [(xmin + xmax) / 2 * w, (ymin + ymax) / 2 * h, (xmax - xmin) * w, (ymax - ymin) * h]
            )
            labels.append(int(label))
            scores.append(float(score))

    if len(boxes) == 0:
        boxes = np.array([]).reshape(0, 4)
        scores = np.array([])
        labels = np.array([])
    return np.array(boxes), np.array(scores), np.array(labels)


def draw_boxes(img, bbox, identities=None):
    """
    Draw bounding box in original image

    Parameters
    ----------
    img: original image
    bbox: coordinate of bounding box
    identities: identities IDs
    """
    for i, box in enumerate(bbox):
        x1, y1, x2, y2 = [int(i) for i in box]
        # box text and bar
        id = int(identities[i]) if identities is not None else 0
        color = compute_color_for_labels(id)
        label = '{}{:d}'.format("", id)
        t_size = cv2.getTextSize(label, cv2.FONT_HERSHEY_PLAIN, 2, 2)[0]
        cv2.rectangle(img, (x1, y1), (x2, y2), color, 2)
        cv2.rectangle(
            img, (x1, y1), (x1 + t_size[0] + 3, y1 + t_size[1] + 4), color, -1)
        cv2.putText(
            img,
            label,
            (x1, y1 + t_size[1] + 4),
            cv2.FONT_HERSHEY_PLAIN,
            1.6,
            [255, 255, 255],
            2
        )
    return img


def cosin_metric(x1, x2):
    """
    Calculate the consin distance of two vector

    Parameters
    ----------
    x1, x2: input vectors
    """
    return np.dot(x1, x2) / (np.linalg.norm(x1) * np.linalg.norm(x2))

Test person reidentification model

The reidentification network outputs a blob with the (1, 256) shape named reid_embedding, which can be compared with other descriptors using the cosine distance.

base_file_link = 'https://storage.openvinotoolkit.org/repositories/openvino_notebooks/data/data/image/person_'
image_indices = ['1_1.png', '1_2.png', '2_1.png']
image_paths = [utils.download_file(base_file_link + image_index, directory='data') for image_index in image_indices]
image1, image2, image3 = [cv2.cvtColor(cv2.imread(str(image_path)), cv2.COLOR_BGR2RGB) for image_path in image_paths]

# Define titles with images.
data = {"Person 1": image1, "Person 2": image2, "Person 3": image3}

# Create a subplot to visualize images.
fig, axs = plt.subplots(1, len(data.items()), figsize=(5, 5))

# Fill the subplot.
for ax, (name, image) in zip(axs, data.items()):
    ax.axis('off')
    ax.set_title(name)
    ax.imshow(image)

# Display an image.
plt.show(fig)

data/person_1_1.png:   0%|          | 0.00/68.3k [00:00<?, ?B/s]

data/person_1_2.png:   0%|          | 0.00/68.9k [00:00<?, ?B/s]

data/person_2_1.png:   0%|          | 0.00/70.3k [00:00<?, ?B/s]

# Metric parameters
MAX_COSINE_DISTANCE = 0.6  # threshold of matching object
input_data = [image2, image3]
img_batch = batch_preprocess(input_data, extractor.height, extractor.width)
features = extractor.predict(img_batch)
sim = cosin_metric(features[0], features[1])
if sim >= 1 - MAX_COSINE_DISTANCE:
    print(f'Same person (confidence: {sim})')
else:
    print(f'Different person (confidence: {sim})')

Different person (confidence: 0.02726624347269535)

Main Processing Function

Run person tracking on the specified source. Either a webcam feed or a video file.

# Main processing function to run person tracking.
def run_person_tracking(source=0, flip=False, use_popup=False, skip_first_frames=0):
    """
    Main function to run the person tracking:
    1. Create a video player to play with target fps (utils.VideoPlayer).
    2. Prepare a set of frames for person tracking.
    3. Run AI inference for person tracking.
    4. Visualize the results.

    Parameters:
    ----------
        source: The webcam number to feed the video stream with primary webcam set to "0", or the video path.
        flip: To be used by VideoPlayer function for flipping capture image.
        use_popup: False for showing encoded frames over this notebook, True for creating a popup window.
        skip_first_frames: Number of frames to skip at the beginning of the video.
    """
    player = None
    try:
        # Create a video player to play with target fps.
        player = utils.VideoPlayer(
            source=source, size=(700, 450), flip=flip, fps=24, skip_first_frames=skip_first_frames
        )
        # Start capturing.
        player.start()
        if use_popup:
            title = "Press ESC to Exit"
            cv2.namedWindow(
                winname=title, flags=cv2.WINDOW_GUI_NORMAL | cv2.WINDOW_AUTOSIZE
            )

        processing_times = collections.deque()
        while True:
            # Grab the frame.
            frame = player.next()
            if frame is None:
                print("Source ended")
                break
            # If the frame is larger than full HD, reduce size to improve the performance.

            # Resize the image and change dims to fit neural network input.
            h, w = frame.shape[:2]
            input_image = preprocess(frame, detector.height, detector.width)

            # Measure processing time.
            start_time = time.time()
            # Get the results.
            output = detector.predict(input_image)
            stop_time = time.time()
            processing_times.append(stop_time - start_time)
            if len(processing_times) > 200:
                processing_times.popleft()

            _, f_width = frame.shape[:2]
            # Mean processing time [ms].
            processing_time = np.mean(processing_times) * 1100
            fps = 1000 / processing_time

            # Get poses from detection results.
            bbox_xywh, score, label = process_results(h, w, results=output)

            img_crops = []
            for box in bbox_xywh:
                x1, y1, x2, y2 = xywh_to_xyxy(box, h, w)
                img = frame[y1:y2, x1:x2]
                img_crops.append(img)

            # Get reidentification feature of each person.
            if img_crops:
                # preprocess
                img_batch = batch_preprocess(img_crops, extractor.height, extractor.width)
                features = extractor.predict(img_batch)
            else:
                features = np.array([])

            # Wrap the detection and reidentification results together
            bbox_tlwh = xywh_to_tlwh(bbox_xywh)
            detections = [
                Detection(bbox_tlwh[i], features[i])
                for i in range(features.shape[0])
            ]

            # predict the position of tracking target
            tracker.predict()

            # update tracker
            tracker.update(detections)

            # update bbox identities
            outputs = []
            for track in tracker.tracks:
                if not track.is_confirmed() or track.time_since_update > 1:
                    continue
                box = track.to_tlwh()
                x1, y1, x2, y2 = tlwh_to_xyxy(box, h, w)
                track_id = track.track_id
                outputs.append(np.array([x1, y1, x2, y2, track_id], dtype=np.int32))
            if len(outputs) > 0:
                outputs = np.stack(outputs, axis=0)

            # draw box for visualization
            if len(outputs) > 0:
                bbox_tlwh = []
                bbox_xyxy = outputs[:, :4]
                identities = outputs[:, -1]
                frame = draw_boxes(frame, bbox_xyxy, identities)

            cv2.putText(
                img=frame,
                text=f"Inference time: {processing_time:.1f}ms ({fps:.1f} FPS)",
                org=(20, 40),
                fontFace=cv2.FONT_HERSHEY_COMPLEX,
                fontScale=f_width / 1000,
                color=(0, 0, 255),
                thickness=1,
                lineType=cv2.LINE_AA,
            )

            if use_popup:
                cv2.imshow(winname=title, mat=frame)
                key = cv2.waitKey(1)
                # escape = 27
                if key == 27:
                    break
            else:
                # Encode numpy array to jpg.
                _, encoded_img = cv2.imencode(
                    ext=".jpg", img=frame, params=[cv2.IMWRITE_JPEG_QUALITY, 100]
                )
                # Create an IPython image.
                i = display.Image(data=encoded_img)
                # Display the image in this notebook.
                display.clear_output(wait=True)
                display.display(i)

    # ctrl-c
    except KeyboardInterrupt:
        print("Interrupted")
    # any different error
    except RuntimeError as e:
        print(e)
    finally:
        if player is not None:
            # Stop capturing.
            player.stop()
        if use_popup:
            cv2.destroyAllWindows()

Run

Before running a new tracking task, we have to reinitialize a Tracker object

NN_BUDGET = 100
MAX_COSINE_DISTANCE = 0.6  # threshold of matching object
metric = NearestNeighborDistanceMetric(
    "cosine", MAX_COSINE_DISTANCE, NN_BUDGET
)
tracker = Tracker(
    metric,
    max_iou_distance=0.7,
    max_age=70,
    n_init=3
)

Use a webcam as the video input. By default, the primary webcam is set with source=0. If you have multiple webcams, each one will be assigned a consecutive number starting at 0. Set flip=True when using a front-facing camera. Some web browsers, especially Mozilla Firefox, may cause flickering. If you experience flickering, set use_popup=True.

If you do not have a webcam, you can still run this demo with a video file. Any format supported by OpenCV will work.

USE_WEBCAM = False

cam_id = 0
video_file = 'https://storage.openvinotoolkit.org/repositories/openvino_notebooks/data/data/video/people.mp4'
source = cam_id if USE_WEBCAM else video_file

run_person_tracking(source=source, flip=USE_WEBCAM, use_popup=False)

Source ended