This commit is contained in:
Glenn Jocher 2018-08-26 11:05:13 +02:00
parent be17c9aaf6
commit a1bf591a78
1 changed files with 7 additions and 3 deletions

View File

@ -7,7 +7,7 @@ http://www.ultralytics.com  
# Description # Description
The https://github.com/ultralytics/yolov3 repo contains code to train YOLOv3 on the COCO dataset: https://cocodataset.org/#home. Credit to P.J. Reddie for YOLO (https://pjreddie.com/darknet/yolo/) and to Erik Lindernoren for the pytorch implementation this repo is based on (https://github.com/eriklindernoren/PyTorch-YOLOv3). The https://github.com/ultralytics/yolov3 repo contains code to train YOLOv3 on the COCO dataset: https://cocodataset.org/#home. **Credit to P.J. Reddie for YOLO** (https://pjreddie.com/darknet/yolo/) and to **Erik Lindernoren for the pytorch implementation** this work is based on (https://github.com/eriklindernoren/PyTorch-YOLOv3).
# Requirements # Requirements
@ -17,14 +17,18 @@ Python 3.6 or later with the following `pip3 install -U -r requirements.txt` pac
- `torch` - `torch`
- `opencv-python` - `opencv-python`
# Running # Training
Run `train.py` to begin training. Each epoch trains on 120,000 images from the train and validate sets, and validates on 5000 images in the validation set. An Nvidia GTX 1080 Ti will run about 16 epochs per day. Loss plots for the bounding boxes, objectness and class confidence should appear similar to results shown here (coming soon) Run `train.py` to begin training. Each epoch trains on 120,000 images from the train and validate COCO sets, and tests on 5000 images from the validation set. An Nvidia GTX 1080 Ti will run about 16 epochs per day. Loss plots for the bounding boxes, objectness and class confidence should appear similar to results shown here (coming soon)
![Alt](https://github.com/ultralytics/yolov3/blob/master/data/coco_training_loss.png "training loss") ![Alt](https://github.com/ultralytics/yolov3/blob/master/data/coco_training_loss.png "training loss")
# Inference
Checkpoints will be saved in `/checkpoints` directory. Run `detect.py` to apply trained weights to an image, such as `zidane.jpg` from the `data/samples` folder, shown here. Checkpoints will be saved in `/checkpoints` directory. Run `detect.py` to apply trained weights to an image, such as `zidane.jpg` from the `data/samples` folder, shown here.
![Alt](https://github.com/ultralytics/yolov3/blob/master/data/zidane_result.jpg "example") ![Alt](https://github.com/ultralytics/yolov3/blob/master/data/zidane_result.jpg "example")
# Testing
Run `test.py` to test the latest checkpoint on the 5000 validation images. Joseph Redmon's official YOLOv3 weights produce a mAP of .581 using this method, compared to .579 in his paper. Run `test.py` to test the latest checkpoint on the 5000 validation images. Joseph Redmon's official YOLOv3 weights produce a mAP of .581 using this method, compared to .579 in his paper.
# Contact # Contact