TinyVision
TinyVision is an evolving project focused on designing ultra-lightweight image classification models with minimal parameter counts. The goal is to explore what’s actually necessary for fundamental vision tasks by combining handcrafted feature preprocessing with highly efficient CNN architectures.
Install / Use
/learn @SaptakBhoumik/TinyVisionREADME
🧠 TinyVision: Compact Vision Models with Minimal Parameters
TinyVision is an evolving research project focused on designing ultra-lightweight image classification models with minimal parameter counts. The goal is to explore what’s actually necessary for fundamental vision tasks by combining handcrafted feature preprocessing with highly efficient CNN architectures.
<!-- 📦 **Current Release**: [v2.0.0](https://github.com/SaptakBhoumik/TinyVision/releases/tag/v2.0.0) 🔖 **Zenodo DOI**: [10.5281/zenodo.16467349](https://doi.org/10.5281/zenodo.16467349) 📁 **Latest Results & Code**: See the `cat_vs_dog_classifier/final/v2` directory > ⚠️ This release **does not include a paper**, but focuses on the **codebase**, experiment results, and reproducible training scripts. A deeper analysis and formal documentation will come in future updates. -->I am still writting the report for V3 but you can read the draft for V3 report here (not final yet, will be updated soon).
You can also read the note for my personal thought here. It is not polished but a good refernce point
V3 report coming very soon
🚧 Project Status
- ✅ Cat vs Dog Classification
First completed task using a 25,000-image dataset with filter bank preprocessing + compact CNNs.- Achieved up to 86.87% test accuracy with models under 12.5k parameters
- Several models under 5k parameters reached over 83% accuracy, showcasing strong efficiency-performance trade-offs.
- 📂 Final results and code for this task are in the
cat_vs_dog_classifier/final/v2directory.
- ✅ Cifar10 Classification
Second completed task using the Cifar10 dataset but without the filter bank preprocessing.Just relies on compact CNN architectures.- Best results achieved :-
- 22.11 k parameters model achieved 87.38% accuracy
- 31.15 k parameters model achieved 88.43% accuracy
- And more
- 📂 Final results and code for this task are in the
cifar10_classifier/final/v1directory.
- Best results achieved :-
🧪 What's Coming Next
- COMING SOON: Quantization experiments
- Add thorough performance analysis of model architectures to understand why something works while others don't
- Explore new vision tasks (edge detection, object detection, etc.) with compact models
- Expand documentation, architecture diagrams, and visualizations
- Log and reflect on failed or inconclusive experiments critical for understanding design boundaries
🤝 Contributing
This project is currently personal and tracks my ongoing experiments.
I’m not accepting pull requests, but you're welcome to:
- 📬 Open an issue for discussion or feedback
- 💌 Reach me at:
saptakbhoumik.acad@gmail.com - 📢 Follow me on X
💡 Philosophy
Small models aren't just about speed—they’re a design challenge.
How much can we cut before it breaks? What’s essential? What’s fluff?
TinyVision is my attempt to find those answers.
