S

semantic segmentation

Cognitive Orchestration Stack (COS) is an open-source, modular AI infrastructure framework designed to improve efficiency and reduce compute waste by separating reasoning from generation. Instead of relying on monolithic language model calls, COS introduces structured modules for intent understanding, semantic processing, adaptive routing, memory, and synthesis, all governed by strict schema validation through the Spec Hygiene system. The result is a system that prioritizes structured thinking, reuse of reasoning, and intelligent model selection to enable more scalable and transparent AI systems.

This project leverages the Brain Tumor Image database to develop a semantic segmentation model for detecting brain tumors. The model is based on an encoder-decoder U-Net architecture, which classifies each pixel in a medical image as either tumor or healthy tissue. The resulting segmentation is visualized as a binary mask, clearly delineating the tumor region from the surrounding healthy areas.

The model successfully localizes the tumor’s global position with high accuracy but the precise shape and boundaries remain approximate (moderate Dice coefficient). Extending the training phase with additional epochs could refine the segmentation quality, leading to a higher Dice coefficient and more accurate tumor delineation.

An AI-driven pipeline for detecting and quantifying cancer and inflammatory tissue on biopsy slides to support standardized diagnostics. This independent fork focuses on advanced ensembling (SMP <-> nnU-Net), probability fusion, and Test-Time Augmentation (TTA), providing a highly reproducible training and inference workflow.

This is a project to train, use and analyze 2D and 3D neural networks for segmentation.

This is a project to train, use and analyze 2D and 3D neural networks for segmentation. It contains a UI and is implemented in pytorch and django as backend.

[SOICT2023] Official Implementation of MCLDA: Multi-level Contrastive Learning for Domain Adaptive Semantic Segmentation

U-Net Adaptive Generalized Image Binarization for Documents

Mirrored from: https://github.com/venkatakolagotla/robin Originally forked from: https://github.com/masyagin1998/robin

This repository contains all my work related to the study of effectiveness of wavelet feature extraction on: Pose estimation Human segmentation Object detection Image Processing

Dense Prediction by Means of Self-attention Layers - research of models for dense prediction (semantic segmentation) - primarily transformers (models in focus: Segmenter, Swin transformer) and comparison with convolutional models (model in focus: pyramidal SwiftNet). Also, research, design and implementation of pyramidal models of transformer-convolutional model (Segmenter-SwiftNet) and transformer-transformer (Segmenter-Segmenter) type. Implementation is in PyTorch deep learning framework. My graduate thesis computer vision project.