Add Hugging Face Spaces deployment files

Author: muntasir-islam

modalx_v2/spaces/DEPLOY.md

@@ -0,0 +1,92 @@
+# Deploying ModalX v2 on Hugging Face Spaces (FREE GPU)
+
+## Quick Deploy Steps
+
+### 1. Create a Space
+
+1. Go to [huggingface.co/spaces](https://huggingface.co/spaces)
+2. Click **"Create new Space"**
+3. Settings:
+   - **Name:** `modalx-v2`
+   - **SDK:** `Streamlit`
+   - **Hardware:** `T4 small` (FREE GPU!)
+   - **Visibility:** Public or Private
+
+### 2. Upload Files
+
+Upload these files from `modalx_v2/` to your Space:
+
+```
+app.py
+backend.py
+requirements.txt (use spaces/requirements.txt)
+README.md (use spaces/README.md for HF format)
+models/
+  ├── __init__.py
+  ├── transformer_emotion.py
+  ├── action_unit_detector.py
+  ├── gesture_stgcn.py
+  ├── prosody_analyzer.py
+  ├── content_bert.py
+  └── slide_vit.py
+weights/
+  └── (your trained .pt files)
+```
+
+### 3. Using Git
+
+```bash
+# Clone your space
+git clone https://huggingface.co/spaces/YOUR_USERNAME/modalx-v2
+cd modalx-v2
+
+# Copy files
+cp -r /path/to/modalx_v2/* .
+cp spaces/README.md .
+cp spaces/requirements.txt .
+
+# Push
+git add .
+git commit -m "Initial deploy"
+git push
+```
+
+### 4. Wait for Build
+
+The Space will build automatically. Takes ~5-10 minutes.
+
+---
+
+## Free GPU Limits
+
+| Feature | Limit |
+|---------|-------|
+| GPU | T4 (16GB VRAM) |
+| RAM | 16GB |
+| Storage | 50GB |
+| Timeout | 15 min idle |
+| Cost | **FREE** |
+
+---
+
+## Troubleshooting
+
+### Out of Memory
+Use `whisper.load_model("tiny")` instead of "base"
+
+### Slow Cold Start
+First load takes ~2 minutes to download models
+
+### Import Errors
+Check that `protobuf>=3.20.0,<5.0.0` is in requirements.txt
+
+---
+
+## Your Space URL
+
+After deployment:
+```
+https://huggingface.co/spaces/YOUR_USERNAME/modalx-v2
+```
+
+Share this link for the competition demo!

modalx_v2/spaces/README.md

@@ -0,0 +1,36 @@
+---
+title: ModalX v2.0
+emoji: 🧠
+colorFrom: purple
+colorTo: blue
+sdk: streamlit
+sdk_version: 1.28.0
+app_file: app.py
+pinned: false
+license: mit
+hardware: t4-small
+---
+
+# ModalX v2.0 - Deep Learning Presentation Grader
+
+AI-powered presentation analysis using 6 deep learning models:
+
+- 🎭 **Emotion**: Transformer + Multi-Head Attention
+- 👤 **Facial AU**: ResNet-50 + LSTM
+- 🙌 **Gesture**: ST-GCN Network
+- 🎙️ **Prosody**: CNN-BiLSTM
+- 📝 **Content**: DistilBERT
+- 🖼️ **Slides**: Vision Transformer
+
+## Usage
+
+1. Enter student name and ID
+2. Upload video or paste YouTube/Drive URL
+3. Click "Analyze"
+4. Download PDF report
+
+## Competition
+
+ModalX-AI Challenge @ Daffodil International University
+
+**Team:** NL Circuits

modalx_v2/spaces/requirements.txt

@@ -0,0 +1,32 @@
+# Hugging Face Spaces Requirements
+# GPU-enabled deployment
+
+streamlit>=1.28.0
+torch>=2.1.0
+torchvision>=0.16.0
+torchaudio>=2.1.0
+transformers>=4.35.0
+tokenizers>=0.14.0
+sentencepiece>=0.1.99
+librosa>=0.10.1
+openai-whisper>=20231117
+soundfile>=0.12.1
+scipy>=1.11.0
+opencv-python-headless>=4.8.0
+mediapipe>=0.10.7
+Pillow>=10.0.0
+timm>=0.9.12
+einops>=0.7.0
+moviepy>=1.0.3
+yt-dlp>=2023.11.16
+gdown>=4.7.1
+pytesseract>=0.3.10
+fpdf2>=2.7.6
+numpy>=1.24.0
+pandas>=2.1.0
+scikit-learn>=1.3.0
+plotly>=5.18.0
+matplotlib>=3.8.0
+tqdm>=4.66.0
+requests>=2.31.0
+protobuf>=3.20.0,<5.0.0