README.md

metadata

title: Cifar10 Resnet
emoji: 🐢
colorFrom: yellow
colorTo: green
sdk: gradio
sdk_version: 3.39.0
app_file: app.py
pinned: false
license: mit

CIFAR10 Image Classification!!

This space is being used to host a pretrained custom ResNet type of CNN architecture. The model was trained on CIFAR-10 dataset. CIFAR-10 dataset has the following classes in it: 'airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck'.

This repository contains the following files

---

cifar10_resnet_epochs30.ckpt

This is the checkpoint file that was saved post training of 30 epochs. PytorchLightning framework has been used to save the model.

app.py

This module contains the inference function that takes the user inputs such as image, show_grad, layers, and number of predictions and outputs the prediction as required.

helper_fn.py

Module contains the required helper function such as loading the CustomResnet2 model with the provided checkpoint file, and map_location (CPU/cuda). Also contains the train and test image augmentations that were applied on CIFAR10 dataset.

---

Inputs from the User:

The user is allowed to input the following:

1. upload the image
2. A checkbox to take user input on whether to display GradCAM image
3. A slider to select the layer from which GradCAM image can be extracted.
4. A slider to select the total number of classes to be displayed.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference