A collection of small datasets and pretrained models that you can use with PyLabel or any way you want to.
The squirrels and nuts dataset contains images of squirrels coming to a window to feed on nuts in Seattle, Washington. The images were captured using the Shinobi open source CCTV solution. Then the frames were extracted and labelled using AWS SageMaker Ground Truth.
- squirrelsandnuts_train.zip 100 images with Yolov5 annotations.
- squirrelsandnuts_yolov5.pt is a YOLOv5 model trained on the squirrelsandnuts dataset for 100 epochs with a mAP@.5 of 0.95. You can use these commands to download and use the model.
!wget https://github.com/pylabel-project/datasets_models/blob/main/squirrelsandnuts/squirrelsandnuts.pt?raw=true -O squirrelsandnuts.pt
model = torch.hub.load('ultralytics/yolov5', 'custom', path='squirrelsandnuts.pt', force_reload=True)
BCCD Dataset is a small-scale dataset for blood cells detection. The official repo for this dataset is here https://github.com/Shenggan/BCCD_Dataset.
- BCCD_coco.zip is a copy of the full dataset with annotations in COCO format. The orginal dataset uses VOC format.
- BCCD_yolov5.pt is a YOLOv5 model trained on the BCCD dataset for 20 epochs with a mAP@.5 of 0.91. You can use these commands to download and use the model.
!wget https://github.com/pylabel-project/datasets_models/blob/main/BCCD/BCCD_yolov5.pt?raw=true -O BCCD_yolov5.pt
model = torch.hub.load('ultralytics/yolov5', 'custom', path='BCCD_yolov5.pt', force_reload=True)
Road Sign dataset is a small version of the Kaggle dataset that contains 877 images of 4 distinct classes for the objective of road sign detection. The version in this repo contains 43 images and bounding box annotations. These are normally provided in the PASCAL VOC format, but have been converted to the YOLO format using our package. https://www.kaggle.com/andrewmvd/road-sign-detection