This repository bases on Ultralytics open-source YOLO repository https://github.com/ultralytics/yolov5. changes model architecture such as replacing upsample with transpose, setting upsample's ceil_mode to True in order to convert to caffe correctly and deploy on edge device.

Pretrained Checkpoints

Model	APval	APtest	AP50	SpeedGPU	FPSGPU		params	FLOPs
YOLOv5s-origin(ckpt)	35.5	35.5	55.0	2.1ms	476		7.1M	12.6B
YOLOv5s-ours(ckpt)	34.9	-	54.6	2.1ms	476		7.1M	12.6B

** AP^test denotes COCO test-dev2017 server results, all other AP results in the table denote val2017 accuracy.
** All AP numbers are for single-model single-scale without ensemble or test-time augmentation. Reproduce by python test.py --img 736 --conf 0.001
** Speed_GPU measures end-to-end time per image averaged over 5000 COCO val2017 images using a GCP n1-standard-16 instance with one V100 GPU, and includes image preprocessing, PyTorch FP16 image inference at --batch-size 32 --img-size 640, postprocessing and NMS. Average NMS time included in this chart is 1-2ms/img. Reproduce by python test.py --img 640 --conf 0.1
** All checkpoints are trained to 300 epochs with default settings and hyperparameters (no autoaugmentation).

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

Inference

Downloading YOLOv5s-ours (https://drive.google.com/file/d/1F6NFPNFC_IMTb3JjS8X-2v3nFsPRh_oB/view?usp=sharing) as yolov5s.pt.

Inference can be run on most common media formats. Results are saved to ./inference/output.

$ python detect.py --source file.jpg  # image 
                            file.mp4  # video
                            ./dir  # directory
                            0  # webcam
                            rtsp://170.93.143.139/rtplive/470011e600ef003a004ee33696235daa  # rtsp stream
                            http://112.50.243.8/PLTV/88888888/224/3221225900/1.m3u8  # http stream

To run inference on examples in the ./inference/images folder:

$ python detect.py --source ./inference/images/ --weights yolov5s.pt --conf 0.4

Reproduce Our Training

Download COCO, install Apex and run command below. Training times for YOLOv5s/m/l/x are 2/4/6/8 days on a single V100 (multi-GPU times faster). Use the largest --batch-size your GPU allows (batch sizes shown for 16 GB devices).

$ python train.py --data coco.yaml --cfg yolov5s.yaml --batch-size 196