You can deploy this model and start earning money today!

neelimapreeti297/
classification

Name: classification
Availability: InStock

keras

image-classification

mit

Views

No views yet

Model Name

panda_cat_dog_classification

Model Description

This model classifies animals among pandas, cats and dogs. It was trained using custom CNN model.

Developed by: Neelima Monjusha Preeti
Model type: custom CNN model
Language(s): Python
License: MIT
Contact: monjusha.2017@juniv.edu

Task Description

This panda_cat_dog_classification app classifies between panda, cat, or dog. So the input field is going to take input an image of one of three classes of dog, cat and panda.Then as output, it is going to show the name of the animal to which it belongs. It first processes the data and resizes it. Then custom CNN model is developed. The loss function and optimizer are calculated. After that, the custom model is trained and tested then the app is launched using gradio in Hugging Face.

Data Preprocessing

The image dataset is preprocessed with the following portion:

bash

transform = transforms.Compose([
  transforms.Resize((224,224)),
  transforms.ToTensor(),
  transforms.Normalize((0.485,0.456,0.406),(0.229,0.224,0.225))
  ])

transforms.Resize((224,224)) resizes the input image to (224, 224) pixels. transforms.ToTensor() converts the input image into a PyTorch tensor. Neural networks typically operate on tensors, so this transformation converts the image into a format suitable for further processing. transforms.Normalize(()) normalizes the tensor image with mean and standard deviation. The values provided are mean and standard deviation values for each channel in the tensor.

Model Architecture

The model was trained with custom CNN() model. this CNN architecture consists of two convolutional layers followed by two fully connected layers, and it is designed for a classification task with three classes.

bash

class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        self.conv1 = nn.Conv2d(3, 6, 5)
        self.conv2 = nn.Conv2d(6, 16, 5)
        self.pool = nn.MaxPool2d(2, 2)
        self.fc1 = nn.Linear(16 * 53 * 53, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 3)

    def forward(self, x):
        x = self.conv1(x)
        x = self.pool(x)
        x = self.conv2(x)
        x = self.pool(x)
        x = x.view(-1, 16 * 53 * 53)
        x = self.fc1(x)
        x = self.fc2(x)
        x = self.fc3(x)
        return x

Then used batch_size = 8 and CrossEntropyLoss() for loss function. Then used Adam optimizer with a learning rate 0.001 for optimization process.

bash

loss_function = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

Training Loop

Loading the data then breaking it into mini batches. Then forward pass and loss function calculation. After that backward propagation and optimization. Backward Propagation and Optimization:

bash

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

Test data

Test data loaded and calculate the accuracy.

The accuracy was 53.333333333333336% .

Result Analysis

The packages needed for creating the huggingface interface is loaded with:

bash

import gradio as gr
import torch
from torchvision import transforms

The model was saved with the following:

bash

model_scripted = torch.jit.script(model)
model_scripted.save('./models/cat_dog_cnn.pt')

HuggingFace Result analysis

First the custom model cat_dog_cnn.pt is loaded. Then the output function is specified. As this is a Image Classification model.

bash

|---app_data
|      |---cat.jpg
|      |---dog.jpg
|      |---panda.jpg
|

Example images are loaded. The classes for prediction are - CLASSES = ["Cat", "Dog", "Panda"]. The output function for prediction is

bash

def classify_image(inp):
  inp = transform(inp).unsqueeze(0)
  out = model(inp)
  return CLASSES[out.argmax().item()]

This will return the classes of the input image.

Interface Creation

For creating huggingface interface this following portion is added:

bash

iface = gr.Interface(fn=classify_image,
                     inputs=gr.Image(type="pil", label="Input Image"),
                     outputs="text",
                     examples=[

                       "./app_data/cat.jpg",
                       "./app_data/dog.jpg",
                       "./app_data/panda.jpg",
                     
                               
                     ])

This portion is going to create an interface for taking the image input. Then example images and output is defined to be the classes from cat, dog and panda. Now with the following the interface of the app is loaded.

iface.launch()

The app interface looks like this:

Project Structure

bash

|
|---app_data 
|       |---images(used for examples)
|
|---models
|       |---cat_dog_cnn.pt
|
|---train(image dataset for training)
|
|---test(image dataset for testing)
|
|---Readme.md(about project)
|
|---app.py(the interface for project)
|
|---requirements.txt(libraries needed for project)
|
|---main.ipynb(project code)

How to Run

bash

git clone https://huggingface.co/spaces/neelimapreeti297/panda_cat_dog_classification/tree/main

cd panda_cat_dog_classification

pip install -r requirements.txt

python app.py

License

This project is licensed under the MIT License.

Contributor

Neelima Monjusha Preeti - monjusha.stu2017@juniv.edu

link: https://huggingface.co/spaces/neelimapreeti297/panda_cat_dog_classification

neelimapreeti297/classification

Model Name

Model Description

Task Description

Data Preprocessing

Model Architecture

Training Loop

Test data

Result Analysis

HuggingFace Result analysis

Interface Creation

Project Structure

How to Run

License

Contributor

neelimapreeti297/
classification