🎯
SpotSeg — Text-Guided Object Segmentation
A multimodal vision tool that highlights any object in an image using natural language. Type "dog" and the dog lights up. Combines CLIPSeg for zero-shot segmentation with YOLOv8 for automatic multi-object detection — validated in resource-constrained environments.
CLIPZero-Shot LearningYOLOv8Image SegmentationUltralytics
🔬
SummaScope — AI Document Analyzer
An end-to-end NLP pipeline that extracts abstractive summaries, named entities, and key phrases from any document. Combines DistilBART, BERT-NER, and YAKE into a single analysis tool deployed on Hugging Face Spaces.
TransformersBARTBERTGradioHugging Face
🌀
DepthLens — Monocular Depth Estimation
A computer vision tool that estimates per-pixel depth from a single image using MiDaS. Upload any photo and get a detailed depth map — no stereo cameras or LiDAR required. Supports multiple colormaps, side-by-side comparisons, and overlay modes.
PyTorchOpenCVMiDaSGradioComputer Vision
🛠️
Tooloque — Free Online Toolkit
A collection of 30+ fast, free, privacy-first online tools — calculators, unit converters, text utilities, developer tools, and CSS generators. Everything runs client-side in the browser with no sign-up required. Built with React and Vite.
ReactViteJavaScriptSEOResponsive Design
Visit Tooloque.com →
🧠
WikiLM — Wikipedia Language Model
A GPT-style transformer language model trained from scratch on Wikipedia text data. Includes a custom BPE tokenizer, configurable model architectures (25M–124M params), built entirely in PyTorch with no pre-trained weights.
Transformer ArchitectureBPE TokenizationNumPyHuggingFace Datasets
View on GitHub →
📸
Image Deblurring with Deep Learning
A complete image restoration pipeline that recovers sharp images from blurred photos using a U-Net encoder-decoder architecture. Trained on the DIV2K dataset with realistic motion and focus blur augmentations, evaluated with PSNR and SSIM metrics.
U-NetTensorBoardOpenCVDeep LearningImage Processing
View on GitHub →