TrADe Re-ID – Improving Person Re-Identification using Tracking and Anomaly Detection

---

TrADe is a new live RE-ID approach to generate lower high-quality galllery. TrADe first uses a Tracking algorithm to generated a tracklets. Following, an Anomaly detection model is used to select a best representative of each tracklet.

This repository implements a live RE-ID approach and testing procedure from this paper.

Abstract

---

Person Re-Identification (Re-ID) is a computer vision problem, which goal is to search for a person of interest (query) in a network of cameras. In the classic Re-ID setting the query is sought in a curated gallery containing properly cropped images of entire human bodies. Recently, the live Re-ID setting was introduced to represent better the practical application context of Re-ID. It consists in searching for the query in short videos, containing whole scene frames. The initial baseline proposed to address live Re-ID used a pedestrian detector to build a large search gallery from the video, and applied a classic Re-ID model to find the query in the gallery. However, the galleries generated were too large and contained low-quality images, which decreased the live Re-ID performance significantly. Here, we present a new live Re-ID approach called TrADe, to generate lower high-quality galleries. TrADe first uses a Tracking algorithm to identify tracklets (sequence of images of the same individual) in the gallery. Following, an Anomaly Detection model is used to select a single representative of each tracklet. We validate the efficiency of TrADe on the live Re-ID version of the PRID-2011 dataset and show significant improvements over the initial baseline.

General Requirements:

---

1. torch
2. tensorflow-gpu
3. keras
4. pyqt5
5. opencv
6. CUDA &&cuDNN.

Donwload weight

• Download TrADe's weights here.

Setup

Here are sample steps for setup over Ubuntu-20.04. You must install the follow:

• CUDA >= 11.4.
• nvidia-docker
• docker compose plugin

---

• Please check nvidia-docker with the next step.

  nvidia-docker run --rm nvidia/cuda:11.0.3-cudnn8-runtime-ubuntu20.04 nvidia-smi

  We must see a console similar below.

foo@bar:~$ 
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.141.03   Driver Version: 470.141.03   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro P5000        Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   59C    P5    12W /  N/A |    543MiB / 16278MiB |     15%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

---

• Please check docker compose is installed correctly:

  docker compose version

  We must see a console similar below.

foo@bar:~$ 
Docker Compose version v2.xx.x

---

• Download docker image.

  docker pull luigymach/trade_dev:1.3.0 

  We must see a console similar below.

foo@bar:~$  
pull luigymach/trade_dev:1.3.0
1.3.0: Pulling from luigymach/trade_dev
Digest: sha256:ccdf653c2a8f32a5390f1270d7df437a12f65fb12f9f5b2408e809a66d8a6bbc
Status: Image is up to date for luigymach/trade_dev:1.3.0
docker.io/luigymach/trade_dev:1.3.0

TL;DR

Docker compose

---

• To spin-up a container

  docker compose --env-file ./docker/.env.trade up --detach

  We must see a console similar below.

foo@bar:~$  
[+] Running 4/4
 ⠿ Network trade_default     Created             0.0s
 ⠿ Container trade_notebook  Started             1.3s
 ⠿ Container trade_dev       Started             1.2s
 ⠿ Container trade_base      Started             1.2s

---

• to execute ''run_TrADe.py'', ''eval_TrADe.py'', etc.

   docker compose --env-file ./docker/.env.trade exec trade_dev bash

  We must see a console similar below.

docker@yyy:~$ ls
TrADe  data
docker@yyy:~$ cd TrADe/
docker@yyy:~/TrADe$ python <file>.py

---

• Please enter the below link if you want to execute some jupyter notebook.

  http://localhost:8888/

---

• To down services of docker compose

  docker compose --env-file ./docker/.env.trade down

  We must see a console similar below.

foo@bar:~$
[+] Running 4/4
 ⠿ Container trade_dev       Removed             0.5s
 ⠿ Container trade_notebook  Removed             0.5s
 ⠿ Container trade_base      Removed             0.5s
 ⠿ Network trade_default     Removed             0.1s