Yolov5-Vehicle-Counting

A computer vision and artificial intelligence project to detect and counting vehicles.

Description

This project is using YOLOV5 and Deep Sort Algorithm to perform object recognition and tracking realtime.

Dependencies

• OS: Windows/ Linux
• Python >= 3.8

pip install -r requirement.txt

Installing

• (Optional) Download pretrained weight vehicle_weight.pt

Executing track.py to track and count object

1. Media source

$ python track.py --source 0  # webcam
                           img.jpg  # image
                           vid.mp4  # video
                           path/  # whole directory
                           'https://youtu.be/link'  # YouTube
                           'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream

2. Yolo Model

$ python track.py --source 0 --yolo_model yolov5n.pt 
                                          yolov5s.pt 
                                          yolov5m.pt       
                                          yolov5l.pt 
                                          yolov5x.pt 
                                          ...

3. Image size (pixels)

$ python track.py --source 0 --yolo_model yolov5n.pt --img 640  
                                                     --img 1280 
                       

4. How to save output

$ python track.py --source 0 --yolo_model yolov5n.pt --img 640  --save-vid
                                                                --show-vid
                                                                --save-txt

Train your own weight with custom data

1. Change to yolov5 directory

$ cd yolov5

2. Prepare your custom data in /datasets directory

create datasets/data.yaml

$ cd datasets
$ nano data.yaml

# example of data.yaml
names:
- Car
- Motorcycle
- Truck
- Bus
- Bicycle
nc: 5                           #no. of classes
train: dataset/train/images     # path to train images directory
val: dataset/valid/images       # path to validation images directory

create train and validation directory

$ mkdir train valid
$ mkdir train/images train/labels valid/images valid/labels

check the directories are existed

$ ls */

# the output should be look like this
# 
# train/:
# images labels labels.cache
#
# valid/:
# images labels labels.cache

3. Put dataset in /datasets/train and /datasets/valid directories

• (Optional)download vehicles datasets from MaryamBoneh/Vehicle-Detection by this link

4. Edit yolo config in /models or download from https://github.com/ultralytics/yolov5/tree/master/models

$ nano /models/yolov5.yaml

# Example in case of using yolov5m.yaml -> change no. of classes in parameters section to match your classes. Like below!
# Parameters
nc: 5 #  <--- this one (number of classes)
depth_multiple: 0.67  # model depth multiple
width_multiple: 0.75  # layer channel multiple
anchors:
  - [10,13, 16,30, 33,23]  # P3/8
  - [30,61, 62,45, 59,119]  # P4/16
  - [116,90, 156,198, 373,326]  # P5/32
                  .
                  .
                  .

5. Run train.py (ex. img_size = 640px, batch_size = 16, epochs = 3)

For new train weight

$ python train.py --img 640 --batch 16 --epochs 48 --data models/custom_yolov5.yaml --weights '' --cache

For continue trained weight (trained weight name = 'custom_yolov5.pt')

$ python train.py --img 640 --batch 16 --epochs 48 --data models/custom_yolov5.yaml --weights custom_yolov5.pt --cache

Acknowledgments

Ultralytics (Yolov5)

Mikel Broström (Yolov5 + Deep Sort with OSNet)

MaryamBoneh (Vehicles Dataset)

Contributor

• Ridhirin Lukkanawaraporn
• Siriphak Amrapal
• Charnkanit Keawwong