Open to Work

1 3 13

Eric Chung PRO

DawnC

AI & ML interests

Computer Vision, LLM, Hybrid Architectures, MultiModel, Reinforcement Learning

Organizations

None yet

posted an update about 2 months ago

Post

2898

VividFlow: Complete AI Image Transformation Platform 🎬🎨✨
Three powerful creative tools in one streamlined workspace. VividFlow combines professional video generation, intelligent background replacement, and artistic style transfer to transform your images with precision and creativity.

🎭 Triple Creative Powers
- Cinematic Video Generation transforms static images into smooth motion sequences from 0.5 to 5 seconds. Eight curated motion categories cover portraits, products, landscapes, and artistic content with precision-tuned templates.

- Intelligent Background Replacement generates photorealistic scenes from 24 professionally crafted presets spanning studios, natural environments, urban settings, and seasonal atmospheres. Advanced edge refinement handles complex subjects, while the built-in Touch Up tool eliminates artifacts through AI-powered inpainting for flawless results.

- Artistic Style Transfer converts photographs into stunning interpretations across six distinct styles including 3D Cartoon, Anime, Watercolor, and Oil Painting. Five balanced style blends create unique hybrid aesthetics, with optional Face Restore preserving subject identity during transformation.

⚡ Optimized Performance
Video generation completes in approximately 4 minutes with ongoing optimization targeting sub-60-second processing. Background replacement finishes in 30-40 seconds, while style transfer delivers results in 20-30 seconds. The independent three-tab architecture ensures smooth workflow without performance conflicts.

🎯 Professional Control
Seed-based reproducibility guarantees consistent results across all features. Background generation offers flexible composition modes, adjustable edge softening, and instant mask preview. Comprehensive parameter controls enable precise creative direction.

👉 Try it now: DawnC/VividFlow

Support with a ❤️ — your engagement drives priorities!
#AI #DeepLearning #ImageToVideo #StyleTransfer #CreativeAI

repliedto their post 2 months ago

The system requires substantial VRAM due to its dual-model architecture.

Video generation utilizes the Wan2.2-I2V-A14B model with FP8 quantization, requiring approximately 36GB for model weights plus an additional 4-6GB for inference overhead, bringing the minimum requirement to 40GB VRAM for stable operation.

Background generation employs Stable Diffusion XL alongside OpenCLIP and segmentation models, consuming approximately 14-17GB total with inference overhead included, making 24GB VRAM theoretically sufficient though 28-32GB is recommended for reliability.

The dual-tab architecture ensures only one feature loads at a time, allowing configuration based on your primary use case.

posted an update 2 months ago

Post

2979

VividFlow: AI Image Enhancement & Video Generation 🎬🎨

Bring your images to life with cinematic motion AND create stunning AI backgrounds! VividFlow combines professional-grade video generation with intelligent background replacement in one streamlined platform.

🎭 Dual Creative Powers
Transform any static image into high-quality dynamic videos with smooth, natural motion ranging from 0.5 to 5 seconds. Choose from curated motion templates across 8 categories designed for portraits, products, landscapes, and artistic content. Create photorealistic backgrounds by selecting from 24 professionally crafted scene presets spanning studios, natural environments, urban settings, and artistic atmospheres...etc.

⚡ Optimized Performance
Video generation currently completes in 4-5 minutes with active optimization underway to dramatically reduce processing time. Background replacement finishes in 30-40 seconds after initial loading. The independent dual-tab design ensures smooth workflow without performance conflicts.

🎯 Complete Creative Control
Achieve perfectly consistent results with seed-based reproducibility and adjustable duration for video generation. Background creation offers flexible composition modes, precision edge softening for challenging subjects, and instant mask preview for quality verification.

📈 Continuous Innovation
Ongoing optimization targets significantly faster video generation through advanced model preparation. Future enhancements include expanded template libraries, batch processing capabilities, and industry-specific presets shaped by community feedback.

👉 Try it now: DawnC/VividFlow

Support development with a ❤️ — your engagement shapes future priorities!
#AI #ImageToVideo #BackgroundGeneration #VideoGeneration

2 replies

repliedto their post 3 months ago

Thanks for letting me know! This is likely due to ZeroGPU limitations rather than a bug in the Space.
I’ve just increased the GPU duration from 240s to 300s and restarted the Space to improve stability.
If it still happens, trying again later usually helps. Thanks!

repliedto their post 3 months ago

Thanks, Please stay tuned to this project !

posted an update 3 months ago

Post

2658

VividFlow: AI Image-to-Video Generation 🎬✨

Bring your images to life with cinematic motion! VividFlow transforms any static image—portraits, artwork, products, or landscapes, into dynamic videos with professional animation quality.
The system supports both curated motion templates and custom natural language prompts, giving you complete creative freedom to describe camera movements, subject actions, and atmospheric effects in your own words.

What's Inside?
🎭 Smart Motion Templates — 8 curated categories from fashion cinematography to wildlife animations, each with tested prompts that prevent common artifacts like phantom hands in portraits

⚡ Optimized Engine — Powered by Wan2.2-I2V-A14B with Lightning LoRA distillation and FP8 quantization for memory-efficient inference

🎯 Full Creative Control — Seed-based reproducibility for consistent results, adjustable duration from half a second to five seconds, optional AI prompt expansion with Qwen2.5 for enhanced descriptions, and real-time resolution preview

Current Performance & Development Roadmap
VividFlow runs on ZeroGPU with generation taking about 3-4 minutes for 3-second videos. While I am actively optimizing the pipeline to reduce this time, the current version prioritizes output stability and quality, results are worth the wait!

Future development focuses on dedicated GPU deployment for faster processing, batch generation to create multiple variations at once, and expanding our motion template library based on what the community wants to see.

👉 Try it now: DawnC/VividFlow

If VividFlow brings motion to your creative vision, please show your support with a ❤️, your engagement influences future development priorities!

#AI #ImageToVideo #GenerativeAI #VideoGeneration #DeepLearning

4 replies

repliedto their post 3 months ago

Sorry to hear that. Could you let me know which Indian breed your dog is?
The model is currently trained on 124 specific breeds, so if your dog’s breed isn’t in that list, it won’t be recognized. I’m working on expanding the coverage to include more regional breeds based on user feedback like yours.
Thanks for testing and letting me know.

posted an update 3 months ago

Post

3864

PawMatchAI — Smarter, Safer, and More Thoughtful Recommendations 🐕✨

🐾 Recommendation system update — deeper reasoning, safer decisions
Over the past weeks, user feedback led me to rethink how PawMatchAI handles description-based breed recommendations. Instead of only matching surface-level preferences, the system now implements a multi-dimensional semantic reasoning architecture that emphasizes real-life compatibility and risk awareness.

Key technical improvements:
- SBERT-powered semantic understanding with dynamic weight allocation across six constraint dimensions (space, activity, noise, grooming, experience, family)

- Hierarchical constraint management distinguishing critical safety constraints from flexible preferences, with progressive relaxation when needed

-Multi-head scoring system combining semantic matching (15%), lifestyle compatibility (70%), constraint adherence (10%), and confidence calibration (5%)

-Intelligent risk filtering that applies graduated penalties (-10% to -40%) for genuine incompatibilities while preserving user choice

The goal: 👉 Not just dogs that sound good on paper, but breeds people will actually thrive with long-term.

What's improved?
- 🎯 Clearer separation of must-have safety constraints versus flexible preferences
- 🧠 Bidirectional semantic matching evaluating compatibility from both user and breed perspectives
- 🔍 Context-aware prioritization where critical factors (safety, space, noise) automatically receive higher weighting

What's next?
- 🐕 Expanding behavioral and temperament analysis dimensions
- 🐾 Extension to additional species with transfer learning
- 📱 Mobile-optimized deployment for easier access
- 🧩 Enhanced explainability showing why specific breeds are recommended

👉 Try PawMatchAI: DawnC/PawMatchAI

#AIProduct #SBERT #RecommendationSystems #DeepLearning #MachineLearning #NLP

2 replies

posted an update 3 months ago

Post

5564

Intelligent Inpainting for Precise Creative Control 🎨✨

Transform your images with AI-powered precision! SceneWeaver delivers professional-quality image composition with intelligent background replacement and advanced object manipulation.
What's New in This Update?

🖌️ Object Replacement — Select and transform any element in your scene with natural language prompts while maintaining perfect visual consistency with surrounding content

🗑️ Object Removal — Intelligently remove unwanted objects with context-aware generation that preserves natural lighting, shadows, and scene coherence

🎯 Context-Aware Processing — Advanced inpainting technology ensures seamless integration across all regenerated regions

Core Capabilities
⚡ One-click transformation with smart subject detection, 24 curated professional backgrounds, custom scene generation through text prompts, and studio-quality results powered by BiRefNet, Stable Diffusion XL, and ControlNet Inpainting.

Current Infrastructure & Future Vision
SceneWeaver operates on ZeroGPU with dynamic resource allocation, resulting in extended processing times during peak usage. Based on community demand, I am exploring cloud deployment with dedicated GPU resources for enhanced speed and batch processing capabilities.

Active development focuses on expanding background variety, refining edge quality, and advancing toward intelligent object addition with automatic shadows and reflections—making professional image composition accessible to everyone without technical expertise.

👉 Try it here: DawnC/SceneWeaver

If SceneWeaver helps bring your creative vision to life, please give it a ❤️ — your support influences future development and infrastructure investments!

#AI #Inpainting #DeepLearning #ComputerVision #StableDiffusion #Photography

posted an update 4 months ago

Post

3513

SceneWeaver — AI-Powered Background Generation & Image Composition 🎨✨
Transform ordinary portraits into professional studio shots with just one click!

What can SceneWeaver do?
- 📸 Upload any portrait photo and instantly generate stunning, professional-quality backgrounds

- 🎭 Smart Subject Detection — Automatically identifies and extracts people, pets, or objects from your photos, even handling tricky cases like dark clothing and cartoon characters.

- 🌄 Creative Scene Library — Choose from 24 professionally curated backgrounds spanning offices, nature landscapes, urban settings, artistic styles, and seasonal themes, or describe your own custom vision.

- ⚙️ Professional Results — Delivers studio-quality compositions in seconds, saving hours of manual editing work while maintaining natural lighting and color harmony.

What's next?
🎬 Enhanced context-aware generation
🎨 Batch processing for multiple style variations
🔧 Higher resolution output support
🌐 Accessible cloud deployment

Current Status: Under active development with continuous improvements to edge quality, background variety, and processing efficiency.

My goal: To make professional-quality image composition accessible to everyone, whether you're a photographer needing quick background changes, a content creator building your social media presence, or simply someone who wants their photos to look their absolute best.

👉 Try it here: https: DawnC/SceneWeaver

If SceneWeaver helps bring your creative vision to life, please give it a ❤️ to this project — your support inspires ongoing innovation!

#AI #Photography #ImageEditing #ContentCreation #GenerativeAI #DeepLearning

repliedto their post 4 months ago

Glad u like it !

posted an update 4 months ago

Post

2936

Pixcribe — AI-Powered Social Media Caption Generator 📸✨
Transform your images into compelling stories with intelligent multi-model analysis!

What can Pixcribe do?
📸 Upload photos (up to 10) to get instant AI-generated captions in Traditional Chinese and English

- 🏷️ Brand Recognition — Detects logos and brand elements through visual detection, semantic analysis, and OCR verification.

- 🎨 Scene Understanding — Analyzes composition, lighting conditions, and visual aesthetics to capture your image's mood and context.

- 🔍 Smart Text Extraction — Identifies and incorporates text from your images into captions seamlessly.

- ⚡ Multi-Model Intelligence — Combines YOLOv11 object detection, OpenCLIP semantic understanding, EasyOCR text recognition, U2-Net saliency detection, and Qwen2.5-VL-7B caption generation.

What's next?
🎬 Video processing capabilities
🌐 Enhanced multilingual support
🎯 Interactive caption refinement with user feedback
⚡ Real-time processing optimizations

- Current Status: Under active development — continuously improving brand recognition accuracy and expanding analytical capabilities.

- My goal: To empower content creators, marketers, and social media managers by automating caption generation while maintaining creative quality and cultural authenticity.

👉 Try it here: DawnC/Pixcribe
If you find Pixcribe helpful, please give it a ❤️ , your support drives continuous innovation!

#ComputerVision #VisionLanguageModel #DeepLearning #MachineLearning #ContentCreation #AI #SocialMedia

2 replies

posted an update 7 months ago

Post

6704

PawMatchAI — Now with SBERT-Powered Recommendations! 🐶✨

⭐️ NEW: Description-based recommendations are here!
Just type in your lifestyle or preferences (e.g. “I live in an apartment and want a quiet dog”), and PawMatchAI uses SBERT semantic embeddings to understand your needs and suggest compatible breeds.

What can PawMatchAI do today?
📸 Upload a photo to identify your dog from 124 breeds with detailed info.
⚖️ Compare two breeds side-by-side, from grooming needs to health insights.
📊 Visualize breed traits with radar and comparison charts.
🎨 Try Style Transfer to turn your dog’s photo into anime, watercolor, cyberpunk, and more.

What’s next?
🎯 More fine-tuned recommendations.
📱 Mobile-friendly deployment.
🐾 Expansion to additional species.

My goal:
To make breed discovery not only accurate but also interactive and fun — combining computer vision, semantic understanding, and creativity to help people find their perfect companion.

👉 Try it here:
DawnC/PawMatchAI

If you enjoy PawMatchAI, please give the project a ❤️ — it really helps and keeps me motivated to keep improving!

#ComputerVision #SBERT #DeepLearning #MachineLearning #TechForLife

repliedto their post 8 months ago

Thanks! So glad you enjoyed the technical deep dive.

repliedto their post 8 months ago

Thank you for the kind words! That's a great suggestion, I'll definitely look into it !

posted an update 8 months ago

Post

4528

🎯 Excited to share my comprehensive deep dive into VisionScout's multimodal AI architecture, now published as a three-part series on Towards Data Science!

This isn't just another computer vision project. VisionScout represents a fundamental shift from simple object detection to genuine scene understanding, where four specialized AI models work together to interpret what's actually happening in an image.

🏗️ Part 1: Architecture Foundation
How careful system design transforms independent models into collaborative intelligence through proper layering and coordination strategies.

⚙️ Part 2: Deep Technical Implementation
The five core algorithms powering the system: dynamic weight adjustment, attention mechanisms, statistical methods, lighting analysis, and CLIP's zero-shot learning.

🌍 Part 3: Real-World Validation
Concrete case studies from indoor spaces to cultural landmarks, demonstrating how integrated systems deliver insights no single model could achieve.

What makes this valuable:
The series shows how intelligent orchestration creates emergent capabilities. When YOLOv8, CLIP, Places365, and Llama 3.2 collaborate, the result is genuine scene comprehension beyond simple detection.

⭐️ Try it yourself:
DawnC/VisionScout

Read the complete series:
📖 Part 1: https://towardsdatascience.com/the-art-of-multimodal-ai-system-design/

📖 Part 2: https://towardsdatascience.com/four-ai-minds-in-concert-a-deep-dive-into-multimodal-ai-fusion/

📖 Part 3: https://towardsdatascience.com/scene-understanding-in-action-real-world-validation-of-multimodal-ai-integration/

#AI #DeepLearning #MultimodalAI #ComputerVision #SceneUnderstanding #TechForLife

6 replies

posted an update 9 months ago

Post

3750

🚀 I'm excited to share a recent update to VisionScout, a system built to help machines do more than just detect — but actually understand what’s happening in a scene.

🎯 At its core, VisionScout is about deep scene interpretation.
It combines the sharp detection of YOLOv8, the semantic awareness of CLIP, the environmental grounding of Places365, and the expressive fluency of Llama 3.2.
Together, they deliver more than bounding boxes, they produce rich narratives about layout, lighting, activities, and contextual cues.

🏞️ For example:
- CLIP’s zero-shot capability recognizes cultural landmarks without any task-specific training

- Places365 helps anchor the scene into one of 365 categories, refining lighting interpretation and spatial understanding. It also assists in distinguishing indoor vs. outdoor scenes and enables lighting condition classification such as “sunset”, “sunrise”, or “indoor commercial”

- Llama 3.2 turns structured analysis into human-readable, context-rich descriptions

🎬 So where does video fit in?
While the current video module focuses on structured, statistical analysis, it builds on the same architectural principles as the image pipeline.
This update enables:

- Frame-by-frame object tracking and timeline breakdown

- Confidence-based quality grading

- Aggregated object counts and time-based appearance patterns

These features offer a preview of what’s coming, extending scene reasoning into the temporal domain.

🔧 Curious how it all works?
Try the system here:
DawnC/VisionScout

Explore the source code and technical implementation:
https://github.com/Eric-Chung-0511/Learning-Record/tree/main/Data%20Science%20Projects/VisionScout

🛰️ VisionScout isn’t just about what the machine sees.
It’s about helping it explain — fluently, factually, and meaningfully.

#SceneUnderstanding #ComputerVision #DeepLearning #YOLO #CLIP #Llama3 #Places365 #MultiModal #TechForLife

posted an update 10 months ago

Post

3268

VisionScout Major Update: Enhanced Precision Through Multi-Modal AI Integration

I'm excited to share significant improvements to VisionScout that substantially enhance accuracy and analytical capabilities.

⭐️ Key Enhancements
- CLIP Zero-Shot Landmark Detection: The system now identifies famous landmarks and architectural features without requiring specific training data, expanding scene understanding beyond generic object detection.

- Places365 Environmental Classification: Integration of MIT's Places365 model provides robust scene baseline classification across 365 categories, significantly improving lighting analysis accuracy and overall scene identification precision.

- Enhanced Multi-Modal Fusion: Advanced algorithms now dynamically combine insights from YOLOv8, CLIP, and Places365 to optimize accuracy across diverse scenarios.

- Refined LLM Narratives: Llama 3.2 integration continues to transform analytical data into fluent, contextually rich descriptions while maintaining strict factual accuracy.

🎯 Future Development Focus
Accuracy remains the primary development priority, with ongoing enhancements to multi-modal fusion capabilities. Future work will advance video analysis beyond current object tracking foundations to include comprehensive temporal scene understanding and dynamic narrative generation.

Try it out 👉 DawnC/VisionScout

If you find this update valuable, a Like❤️ or comment means a lot!

#LLM #ComputerVision #MachineLearning #MultiModel #TechForLife

repliedto their post 10 months ago

Glad to hear !

posted an update 10 months ago

Post

2579

🚀 VisionScout Now Speaks More Like Me — Thanks to LLMs!
I'm thrilled to share a major update to VisionScout, my end-to-end vision system.

Beyond robust object detection (YOLOv8) and semantic context (CLIP), VisionScout now features a powerful LLM-based scene narrator (Llama 3.2), improving the clarity, accuracy, and fluidity of scene understanding.

This isn’t about replacing the pipeline , it’s about giving it a better voice. ✨

⭐️ What the LLM Brings
Fluent, Natural Descriptions:
The LLM transforms structured outputs into human-readable narratives.

Smarter Contextual Flow:
It weaves lighting, objects, zones, and insights into a unified story.

Grounded Expression:
Carefully prompt-engineered to stay factual — it enhances, not hallucinates.

Helpful Discrepancy Handling:
When YOLO and CLIP diverge, the LLM adds clarity through reasoning.

VisionScout Still Includes:
🖼️ YOLOv8-based detection (Nano / Medium / XLarge)
📊 Real-time stats & confidence insights
🧠 Scene understanding via multimodal fusion
🎬 Video analysis & object tracking

🎯 My Goal
I built VisionScout to bridge the gap between raw vision data and meaningful understanding.
This latest LLM integration helps the system communicate its insights in a way that’s more accurate, more human, and more useful.

Try it out 👉 DawnC/VisionScout

If you find this update valuable, a Like❤️ or comment means a lot!

#LLM #ComputerVision #MachineLearning #TechForLife

2 replies

Eric Chung PRO

AI & ML interests

Organizations

DawnC's activity