This repository provides state-of-the-art PyTorch implementations for camouflaged object detection using three architectures: IFBO-Net, SegNet, and UNet.

Adjust the num_epochs Accordingly

Lets First of All clone the repository: git clone https://github.com/Sarthak6o1/Camouflaged-Object-Detection-with-IFBO_Net-Model.git

1. Mount Google Drive (for Colab):

   from google.colab import drive drive.mount('/content/drive')

2. Prepare COD10K dataset:

   • Download from Kaggle: https://www.kaggle.com/datasets/getcam/cod10k
   • Unzip into: /content/drive/MyDrive/COD10K/ (or adjust paths in dataset_loader_cod10k.py)

3. Run training and testing:

   # For IFBO-Net python IFBO_Net_model_train.py python IFBO_Net_model_test.py # For SegNet python SegNet_model_train.py python SegNet_model_test.py # For UNet python UNet_model_train.py python UNet_model_test.py

├── IFBO_Net_architecture.py # IFBO-Net model definition ├── IFBO_Net_metrics.py # Metrics for IFBO-Net ├── IFBO_Net_model_train.py # Training pipeline (Adam optimizer) ├── IFBO_Net_model_test.py # Evaluation pipeline ├── IFBO_Net_original.py # Initial implementation ├──SARNet_architecture.py #Implemented the SARNet nodel architecture use the Similar metrics and training ,testing files of SegNet_metrics.py SegNet_model_train.py SegNet_model_test.py ├── SegNet_architecture.py # SegNet definition ├── SegNet_metrics.py # SegNet metrics ├── SegNet_model_train.py # Training pipeline (SGD optimizer) ├── SegNet_model_test.py # Evaluation pipeline ├── SegNet_Original.py # Initial Implementation ├── UNet_architecture.py # UNet definition ├── UNet_metrics.py # UNet metrics ├── UNet_model_train.py # Training pipeline (SGD + scheduler) ├── UNet_model_test.py # Evaluation pipeline ├── UNet_original.py # Initial Implementation ├── dataset_loader_cod10k.py # DataLoader & transforms (Resize 256×256) ├── cod10k_visualizer.py # Initial image visualizer (raw images & masks) ├── Mounting_drive.py # Utility for Google Drive mounting ├── requirements.txt # Python dependencies └── README.md # This file 

• Transforms: All images and masks are resized to (256, 256) and converted to tensors.
• Datasets: COD10KDataset handles reading image-mask pairs from Train and Test folders.
• DataLoader: Batches data (batch_size=8), shuffles train set, uses num_workers=2.

transform = transforms.Compose([ transforms.Resize((256, 256)), transforms.ToTensor() ]) mask_transform = transforms.Compose([ transforms.Resize((256, 256)), transforms.ToTensor() ]) train_loader = DataLoader(...) test_loader = DataLoader(...)

• cod10k_visualizer.py provides an initial image visualization of raw images and corresponding ground-truth masks using matplotlib.
• For prediction heatmaps and advanced plotting, refer to your custom evaluation scripts.

Each training script follows this workflow:

1. Setup: Check for CUDA, instantiate model on device.
2. Optimizer & Scheduler (UNet, SegNet):
   • SGD with momentum (0.9), weight decay (1e-4).
   • StepLR scheduler (step_size=10, gamma=0.1).
   • IFBO-Net uses Adam (lr=0.001).
3. Loss Function:
   • IFBO-Net: BCE loss.
   • SegNet & UNet: BCEDiceLoss (combines BCE + Dice).
4. Training Loop:

   for epoch in range(num_epochs): model.train() for images, masks in train_loader: outputs = model(images) loss = criterion(outputs, masks) optimizer.zero_grad() loss.backward() torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0) optimizer.step() scheduler.step() # if using scheduler

• Loss: BCE, Dice
• Accuracy: Pixel-level accuracy
• MAE: Mean absolute error
• S-measure (Sα): Structural similarity
• E-measure (Eϕ): Enhanced alignment
• Fβw: Weighted F-measure
• Dice Coefficient: Overlap measure
• IoU: Intersection over Union

• COD10K Dataset
• Zhao et al., EGNet: Edge Guidance Network for Salient Object Detection, ICCV 2019.
• Long et al., Fully Convolutional Networks, CVPR 2015.
• Camouflaged object detection with integrated feature fusion and boundary optimization Bin Ge1 · Xiaolong Peng1 · Chenxing Xia1

· Hailong Chen1