Expand object detection to include insects, birds, pests, domestic an…

[anhed0nic]

Expand Object Detection Capabilities

Description

This pull request expands the LLM-Vision-Capabilities project from crop detection to a comprehensive object detection system that can identify a wide range of objects including:

• Crops (existing functionality maintained)
• Insects
• Birds
• Pests
• Domestic animals
• Airplanes
• Cars
• Rocks/Minerals
• Game animals

Changes Made

1. Prompt Updates

• Modified crop_detection.txt and crop_analysis_20250614.txt to identify objects from the expanded categories
• Added conditional logic for different object types (e.g., agricultural fields for crops, N/A for others)

2. Database Schema Changes

• Renamed tables:
  • crop_analysis_results → object_analysis_results
  • crop_detection_results → object_detection_results
• Updated schema to use category and object fields instead of just crop
• Modified all indexes and constraints accordingly

3. Code Modifications

• Updated config.py with new table environment variables
• Modified clickhouse_client.py save functions to handle new JSON structure
• Fixed import paths and relative imports
• Updated embedding generation for semantic search
• Enhanced search functionality in CropSemanticSearch.py (now handles multiple categories)

4. Search and Voice Integration

• Updated semantic search to work with category/object structure
• Modified voice input system to display results for all object types
• Maintained backward compatibility for crop-specific queries

5. Documentation Updates

• Updated README.md to reflect broader scope
• Changed project title to "Voice-Enabled Semantic Object Intelligence"
• Updated environment variable documentation
• Modified example use cases

Technical Details

JSON Response Structure

{
  "category": "crops|insects|birds|pests|domestic animals|Hideo Kojima game characters|airplanes|cars|rocks/minerals|game animals",
  "object": "specific object name",
  "alternate_names": ["..."],
  "color": ["..."],
  "confidence": 0.95,
  "overall_description": "...",
  // ... detailed analysis fields with N/A for irrelevant categories
}

Database Changes

• New tables support all object types with flexible schema
• Semantic search works across all categories
• Embeddings generated for comprehensive text analysis

Testing

• All imports and basic functionality verified
• Database schema updated and compatible
• Search system tested with new structure

Breaking Changes

• Table names changed (requires database migration)
• JSON response structure modified (category/object instead of crop)
• Environment variables updated

Migration Notes

Users will need to:

1. Update database tables using the new SQL files
2. Update environment variables to use CLICKHOUSE_OBJECT_* instead of CLICKHOUSE_CROP_*
3. Re-run any existing crop analysis to populate new fields

This enhancement significantly expands the system's capabilities while maintaining the core vision-language model architecture and search functionality.