https://github.com/girinchutia/fasterrcnn-torchvision-finetuning

Train Torchvision FasterRCNN model with custom COCO dataset
https://github.com/girinchutia/fasterrcnn-torchvision-finetuning

cnn computer-vision deep-learning deep-learning-algorithms deep-neural-networks evaluation faster-rcnn machine-learning object-detection training vision

Last synced: 2 months ago
JSON representation

Train Torchvision FasterRCNN model with custom COCO dataset

• Host: GitHub
• URL: https://github.com/girinchutia/fasterrcnn-torchvision-finetuning
• Owner: GirinChutia
• License: mit
• Created: 2023-07-11T18:19:00.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-07-26T16:56:46.000Z (6 months ago)
• Last Synced: 2024-07-26T18:44:22.955Z (6 months ago)
• Topics: cnn, computer-vision, deep-learning, deep-learning-algorithms, deep-neural-networks, evaluation, faster-rcnn, machine-learning, object-detection, training, vision
• Language: Jupyter Notebook
• Homepage:
• Size: 1.94 MB
• Stars: 8
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Training code for torchvision FasterRCNN model with custom COCO dataset

---

# Faster RCNN :

Faster RCNN is an object detection model introduced in [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://arxiv.org/abs/1506.01497) paper.

The architechure of Faster RCNN model is shown below,

![](utils/image.png)

Faster R-CNN, is composed of two modules. The first module is a deep fully convolutional network that proposes regions, and the second module is the Fast R-CNN detector that uses the proposed regions.

---

# Environment :

- Python version used : 3.9.16

- Create a python or conda environment using ***requirements.txt***

---

# Training Instructions :

To train the Faster RCNN model follow the below steps :

1. Prepare dataset :

    - Prepare dataset in COCO format. It should have the below 2 files & folders

        - Image folder 

        - Annotation file (Json file) in coco format

2. Run :

    > python train.py --epoch 10 --train_image_dir  --val_image_dir  --train_coco_json  --val_coco_json  --batch_size 16 --exp_folder 

    The training weights and tensorboard logs will be saved in experiment folder

    The training and validation logs can be visualized in tensorboard as shown below :

    > Train logs

        ![Alt text](utils/train_logs.png)

    > Val Logs

        ![Alt text](utils/val_logs.png)

---

# Inference :

The instruction about inference with a trained model are discussed in ***demo_inference.ipynb*** notebook