Model Card for CNN Waste Classification (PyTorch & OpenCV)
A PyTorch Convolutional Neural Network (CNN) for multi-class waste classification using images. Predicts 10 types of waste from static images or real-time webcam streams, supporting applications in smart recycling, education, and research. Uses OpenCV for image handling. Trained on the modified Kaggle Garbage Classification dataset.
Model Details
Model Description
A deep learning model for classifying waste into 10 categories: Battery, Cardboard, Clothes, Food Waste, Glass, Metal, Paper, Plastic, Shoes, and Trash. The model uses 6 convolutional layers with batch normalization, dropout, and two fully connected layers. Developed for learning, prototyping, and proof-of-concept smart recycling systems.
- Developed by: Gokul Seetharaman
- Model type: Convolutional Neural Network (CNN)
- License: MIT
- Finetuned from model [optional]: Trained from scratch
Model Sources [optional]
Uses
Direct Use
- Image-based waste detection for smart recycling prototypes
- Educational demonstrations of CNNs, OpenCV, and PyTorch
- Research baselines for waste/material classification
Recommendations
Users should evaluate model performance on their own data and consider retraining or fine-tuning for domain-specific use. It is not recommended to use the model for high-stakes applications without further testing.
How to Get Started with the Model
- Download
best_model.pth and object-detection.py from this repo or GitHub.
- Run
python object-detection.py for webcam or image predictions.
- Modify
object-detection.py to use your own image or video source.
Training Details
Training Data
- Kaggle Garbage Classification dataset
- 10 classes, ~1200 images (split 80/20 train/val)
- Preprocessing: resized to 224x224, normalized, data augmentation (crop, flip, rotation, color jitter, affine)
Training Procedure
- 6 Conv layers, 2 FC layers, dropout, batchnorm
- CrossEntropyLoss, AdamW optimizer, 50 epochs, batch size 8
Preprocessing [optional]
- Images resized to 224x224
- Normalized with ImageNet means/std
- Random data augmentation on train set
Training Hyperparameters
- Training regime: fp32
- Epochs: 50, batch size: 8, optimizer: AdamW, LR: 5e-4
Speeds, Sizes, Times [optional]
- Training time: ~90 minutes on a modern GPU (varies)
- Checkpoint size: ~46MB (
best_model.pth)
Evaluation
Testing Data, Factors & Metrics
Testing Data
- 20% validation split from the Kaggle dataset (stratified)
Factors
- Performance measured per-class (precision, recall, F1-score, support)
Metrics
- Overall accuracy, confusion matrix, precision/recall/F1-score per class
Results
- Validation accuracy: 89.56%
- Most class F1-scores >0.85, with "Plastic" lower due to visual ambiguity
- Full confusion matrix and metrics in GitHub README
Summary
The model reliably classifies 10 types of waste in standard settings. See GitHub for sample images and live demo outputs.
Model Examination [optional]
- No explicit interpretability/visualization methods (e.g., GradCAM) included yet.
Environmental Impact
- Estimated training: <1.5 GPU-hour, carbon footprint minimal for local or single-GPU cloud runs
- Hardware: NVIDIA GeForce GTX 4060 Laptop GPU
- Hours used: ~1.5
Technical Specifications [optional]
Model Architecture and Objective
- See "Model Details" and GitHub repo for the full PyTorch code.
Compute Infrastructure
- Local training with NVIDIA GTX 4060 Laptop GPU, 8GB VRAM, 16GB RAM, Windows 11, Python 3.10
Hardware
- GPU: GTX 4060 (or equivalent, optional CPU)
- RAM: 16GB
Software
- Python 3.10, PyTorch, OpenCV, NumPy
Citation
BibTeX:
1@misc{gokulseetharaman2025wastecnn,
2 title={CNN Waste Classification with OpenCV and PyTorch},
3 author={Gokul Seetharaman},
4 year={2025},
5 url={https://github.com/gokulseetharaman/cnn-waste-classification-opencv-pytorch}
6}
APA:
Gokul Seetharaman. (2025). CNN Waste Classification with OpenCV and PyTorch.
https://github.com/gokulseetharaman/cnn-waste-classification-opencv-pytorch
Model Card Contact
