🧠 ML Model + Flask API (Text & Image Classification)

This project is Flask-based API that uses two open-source machine learning models:

• Text Classification: Classifies text sentiment (positive, negative, neutral) using a pre-trained BERT model.
• Image Classification: Classifies input images using a pre-trained ResNet18 model from PyTorch.

---

🚀 Features

• REST API using Flask
• Two endpoints:
  • /predict-text: accepts raw text in JSON
  • /predict-image: accepts image files (JPG/PNG)
• Returns structured JSON with prediction and confidence
• Works locally on any system with Python installed

---

API Usage

📝 TEXT PREDICTION

URL: POST /predict-text Input: JSON with a "text" field

Example request: json { "text": "I love learning new things!" } Example using Postman or curl: bash

curl.exe -X POST http://127.0.0.1:5000/predict-text -H "Content-Type: application/json" -d "{"text":"I am very happy"}" Example response: json

{ "success": true, "prediction": "positive", "confidence": 0.94 }

🖼️ IMAGE PREDICTION

URL: POST /predict-image Input: multipart/form-data with image field

Example using Postman: Method: POST

URL: http://127.0.0.1:5000/predict-image

Body: form-data

Key: image

Type: File

Value: upload a .jpg or .png image

Example response: json

{ "success": true, "prediction": "Labrador retriever", "confidence": 0.91 }

Model Info

Text Model: 🤗 HuggingFace: distilbert-base-uncased-finetuned-sst-2-english Task: Sentiment analysis Image Model: PyTorch: ResNet18 pretrained on ImageNet Task: Image classification

Requirements

Installed via requirements.txt:

Flask torch torchvision transformers pillow requests

Result JSON Format

Both endpoints return a response like:

{ "success": true, "prediction": "label or list of labels", "confidence": 0.95 }