VisionVault

VisionVault is an AI-powered smart photo gallery that automatically tags your images, lets you search by meaning, generate caption for you, and generate images along with control privacy (private, friends, public), and discover popular photos through a dynamic ranking system based on likes, dislikes, and time.

Models with Image

Object Detection (Bounding boxes around objects)

Models: YOLOv8 Faster R-CNN SSD RetinaNet DETR

Semantic Segmentation (Label every pixel with a class e.g: sky, road, person, etc.)

Models: DeepLabv3+ U-Net PSPNet SegNet HRNet

Instance Segmentation (Detect each object separately + pixel mask per object)

Models: Mask R-CNN YOLACT SOLO DetectoRS

Universal / Promptable Segmentation ((Segment anything, not limited to fixed classes))

Models: Segment Anything Model (SAM) SAM 2

Image Captioning (Generate natural language description)

Models: BLIP-2 Show and Tell OFA ViT-GPT2

Multimodal Vision–Language Models (Can describe, reason, answer questions about images)

Models: CLIP Flamingo LLaVA GPT-4V

Implementation

Detect & Segment Everything

Segment Anything Model (SAM)

Classify Each Segment

CLIP

Generate Caption

BLIP-2
ViT-GPT2

Combination of All of these

LLaVA

Tag Generation: YOlov8

For Caption: Blip 2

Image search: CLIP

Image generation: Stable diffusion