Upload folder using huggingface_hub

README.md

@@ -4,10 +4,11 @@ emoji: 🐦
 colorFrom: green
 colorTo: green
 sdk: gradio
-sdk_version: 4.44.0
+sdk_version: 5.9.1
 app_file: app.py
 pinned: true
 license: mit
+python_version: "3.11"
 tags:
   - bird-identification
   - bioacoustics
@@ -23,15 +24,15 @@ tags:
 
 **Multi-Modal Bird Identification for India**
 
-Competitive with BirdNET | 30 Indian Species | 100% Open Source
+META SAM-Audio Preprocessing | Ollama LLM | 10,000+ Species | CSCR Initiative
 
 ## 🎯 Features
 
-### 🎤 Audio Identification
-- **BirdNET-style analysis** with comprehensive spectral features
-- Mel spectrogram processing
-- Syllable detection and pattern analysis
-- 48kHz processing for high-fidelity analysis
+### 🎤 Audio Identification (with META SAM-Audio)
+- **SAM-Audio style preprocessing** isolates bird calls from noise
+- Text prompts: "bird call", "bird song" for source separation
+- Frequency isolation: 500-10000 Hz bird vocalization range
+- Multi-bird detection via frequency band analysis
 
 ### 📷 Image Identification  
 - Color-based bird matching
@@ -41,50 +42,26 @@ Competitive with BirdNET | 30 Indian Species | 100% Open Source
 ### 📝 Description-Based
 - Natural language bird identification
 - Describe colors, calls, behavior
-- Semantic matching to species database
-
-## 📊 Species Database
-
-30 Indian bird species with comprehensive data:
-
-| Species | Frequency Range | Call Pattern |
-|---------|-----------------|--------------|
-| Asian Koel | 1200-3500 Hz | Ascending whistle |
-| Indian Cuckoo | 600-1500 Hz | Four-note call |
-| Jungle Babbler | 800-4000 Hz | Chattering chorus |
-| Coppersmith Barbet | 1500-3000 Hz | Monotonous tapping |
-| Oriental Magpie-Robin | 2000-7000 Hz | Melodious song |
-| ... and 25 more | | |
+- Semantic matching via LLM
 
 ## 🔬 Technical Details
 
-### Audio Processing Pipeline
-1. **Preprocessing**: 48kHz resampling, bandpass filter (150-15000 Hz)
-2. **Feature Extraction**: 
-   - Spectral: Peak frequency, centroid, bandwidth, flatness, rolloff
-   - Temporal: Syllable detection, onset strength
-   - Pattern: Melodic/repetitive analysis
-3. **Matching**: Multi-feature scoring with quality weighting
+### META SAM-Audio Integration
+```
+Raw Audio → SAM-Audio Preprocessing → Feature Extraction → LLM Identification
+                    ↓
+         Text Prompt: "bird call, bird song"
+         Frequency: 500-10000 Hz
+         Noise Reduction: Spectral gating
+```
 
-### Identification Accuracy
-- Based on BirdNET methodology
-- Frequency matching (40% weight)
-- Syllable rate matching (20% weight)
-- Pattern matching (20% weight)
-- Duration/amplitude (20% weight)
+### LLM Backend
+- **Local**: Ollama with qwen2.5:3b
+- **Cloud**: HuggingFace Inference API (fallback)
 
 ## 🇮🇳 CSCR Initiative
 
-This project is part of the **Citizen Science for Conservation Research** initiative to develop open-source tools for bird researchers in India.
-
-## 🔗 Links
-
-- [GitHub Repository](https://github.com/sohamzycus/eagv2/tree/master/birdsense)
-- [Documentation](https://github.com/sohamzycus/eagv2/blob/master/birdsense/README.md)
-
-## 📝 License
-
-MIT License - Free for research and educational use.
+Part of the **Citizen Science for Conservation Research** initiative for open-source bird identification tools.
 
 ---
 

requirements.txt

@@ -1,8 +1,7 @@
-# BirdSense Pro - HuggingFace Space
-# Use latest stable Gradio
+# BirdSense Pro Requirements
+# Using Gradio 5.x for stable deployment
 
-gradio>=4.0
-numpy
-scipy
-requests
-Pillow
+numpy>=1.21.0
+scipy>=1.7.0
+requests>=2.28.0
+Pillow>=9.0.0