MonFlow

Introduction

MonFlow aims to monitor traffic flow through surveillance cameras. It provides a microservice that receive a static image as input and outputs the number of pedestrians and vehicles that are considered as moving on the picture. Hence, it discards objects that are still, e.g., cars that are parked.

By providing this service, we aim to monitor the influence of the city policy, on COVID-19, in the traffic flow.

Implementation Details

MonFlow is built on top of the excellent UltraLytics’s yolo v3 implementation. Precisely, we customized the original code to train on the surveillance camera images. Furthermore, we used OpenCV to preprocess the incoming images.

As a microservice, it provides a command line interface (CLI) and HTTP interface to interconnect with the reaming system’s components.

Requirements

Python 3.7 or later with all requirements.txt dependencies installed, including torch >= 1.5. To install run:

$ pip install -U -r requirements.txt

Training

Start Training:

$ !python3 train.py --data data/coco_1cls.data --cfg cfg/yolov3-spp.cfg --weights weights/yolov3-spp-ultralytics.pt

Resume Training:

$ !python3 train.py --data data/coco_1cls.data --cfg cfg/yolov3-spp.cfg --weights weights/last.pt --epochs 500

Inference

python3 detect.py --source ...

Image: --source file.jpg
Video: --source file.mp4
Directory: --source dir/
Webcam: --source 0
RTSP stream: --source rtsp://170.93.143.139/rtplive/470011e600ef003a004ee33696235daa
HTTP stream: --source http://112.50.243.8/PLTV/88888888/224/3221225900/1.m3u8

YOLOv3: python3 detect.py --cfg cfg/yolov3.cfg --weights yolov3.pt

Pretrained Checkpoints

Download from: https://drive.google.com/open?id=1LezFG5g3BCW6iYaV89B2i64cqEUZD7e0

Darknet Conversion

$ git clone https://github.com/ultralytics/yolov3 && cd yolov3

# convert darknet cfg/weights to pytorch model
$ python3  -c "from models import *; convert('cfg/yolov3-spp.cfg', 'weights/yolov3-spp.weights')"
Success: converted 'weights/yolov3-spp.weights' to 'weights/yolov3-spp.pt'

# convert cfg/pytorch model to darknet weights
$ python3  -c "from models import *; convert('cfg/yolov3-spp.cfg', 'weights/yolov3-spp.pt')"
Success: converted 'weights/yolov3-spp.pt' to 'weights/yolov3-spp.weights'

marcosroriz/monflow