Arulkumar03/Wheat_HEAD_Detection_Counting_ComputerVision_Model

1# Pointly-Supervised Instance Segmentation2 3Bowen Cheng, Omkar Parkhi, Alexander Kirillov4 5[[`arXiv`](https://arxiv.org/abs/2104.06404)] [[`Project`](https://bowenc0221.github.io/point-sup)] [[`BibTeX`](#CitingPointSup)]6 7<div align="center">8  <img src="https://bowenc0221.github.io/images/cheng2021pointly.png" width="50%" height="50%"/>9</div><br/>10 11## Data preparation12Please follow these steps to prepare your datasets:131. Follow official Detectron2 instruction to prepare COCO dataset. Set up `DETECTRON2_DATASETS` environment variable to the location of your Detectron2 dataset.142. Generate 10-points annotations for COCO by running: `python tools/prepare_coco_point_annotations_without_masks.py 10`15 16## Training17 18To train a model with 8 GPUs run:19```bash20python train_net.py --config-file configs/mask_rcnn_R_50_FPN_3x_point_sup_point_aug_coco.yaml --num-gpus 821```22 23## Evaluation24 25Model evaluation can be done similarly:26```bash27python train_net.py --config-file configs/mask_rcnn_R_50_FPN_3x_point_sup_point_aug_coco.yaml --eval-only MODEL.WEIGHTS /path/to/model_checkpoint28```29 30## <a name="CitingPointSup"></a>Citing Pointly-Supervised Instance Segmentation31 32If you use PointSup, please use the following BibTeX entry.33 34```BibTeX35@article{cheng2021pointly,36  title={Pointly-Supervised Instance Segmentation},37  author={Bowen Cheng and Omkar Parkhi and Alexander Kirillov},38  journal={arXiv},39  year={2021}40}41```42