CNN Feature Map Visualizer

Understand how the CNN is actually making decisions.

This is a local image-recognition + visualization project with:

• FastAPI backend for inference
• React + Vite frontend for interactive visualization
• Trained CNN model at model/final_model.pth

Dataset

The model was trained on CIFAR-10 in:

• Colab Session/imagerecognisitionmodeltraining.ipynb

CIFAR-10 classes:

• airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck

Run Locally (PowerShell)

1) Clone

git clone https://github.com/adityaverma9777/cnn-featuremap-visualizer.git
cd cnn-featuremap-visualizer

2) Start backend + frontend together

From project root:

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
.\start-backend.ps1

The launcher script automatically:

• checks Python and npm
• installs backend dependencies if missing
• installs frontend dependencies if missing
• finds a free backend port (starting from 8000)
• starts backend in one PowerShell window
• starts frontend in another PowerShell window
• auto-connects frontend to the selected backend port

URLs

• Backend URL is printed by launcher (for example http://127.0.0.1:8001 if 8000 is blocked)
• Frontend is usually http://127.0.0.1:5173

Verify It Works

1. Open frontend URL shown by Vite.
2. Upload an image.
3. Run prediction and check output:
   • predicted class
   • confidence
   • class probabilities
   • activation/feature visualization
4. Open backend docs at <backend-url>/docs.

Notes

• This repo is local-only by design.
• No .env setup is required.
• Backend CORS is open for local browser development.