Search - metacurate.io

SORT vs DeepSORT vs ByteTrack vs OC-SORT: Test in Roboflow

The document compares four tracking algorithms (SORT, DeepSORT, ByteTrack, OC-SORT) for multi-object tracking, explains tracking-by-detection, and outlines a workflow using RF-DETR in Roboflow to test and deploy trackers. It covers how each algorithm predicts motion, associates detections, and manages object tracks, ending with a detailed SORT procedure using Kalman filtering and the Hungarian algorithm.

24 mins read Listed 26 days ago in RSS: Roboflow Blog

docs.lightly.ai Documentation

Object Detection - LightlyTrain documentation

LightlyTrain documentation explains how to train and run LTDETR object detection models with YOLO or COCO data. It covers model options, checkpoints, performance figures, and tiled inference for small objects.

8 mins read Listed 5 days ago in reddit/computervision/I measured 44 object detection checkpoints on one RTX 4090. At batch 1, YOLO26-n and YOLO26-x are only 2.7 ms apart end to end

github.com Tools and repositories

GitHub - iMGEai/fotonet: nms-free object detector, Apache-2.0 · GitHub

An open-source, NMS-free object-detection library for Python that supports local inference, training, validation, video tracking, and model export. It provides multiple UltraDense model variants, installation via pip or git, and examples for image and OpenCV frame inference, with training and export workflows including ONNX exports.

2 mins read Listed 23 days ago in reddit/computervision/fotonet v1.1 is out!

huggingface.co Tools and repositories

MTalha2001/LOCUS-2B · Hugging Face

LOCUS-2B grounds referring expressions to image regions using a small box head trained with a Locus recipe. It relies on a cached prefix and Jacobi-verified coordinate tokens, with results reported on standard RefCOCO splits; the repository includes training weights for the head and language-vision merger, while base encoders come from Qwen3-VL-2B-Instruct and are downloaded separately.

3 mins read Listed 18 days ago in reddit/computervision/I made an open-source Jev for object detection 👀

github.com Tools and repositories

GitHub - LeocidDreams/Resnet50FPNYolo26DetectionHead: Experiment to concate Resnet50 backone,FPN neck and Yolo26 Detection Head · GitHub

A PyTorch-based pipeline combines a ResNet50 backbone with an FPN neck and Ultralytics OBB26 head for oriented bounding box detection, using Ultralytics v8 OBB Loss with TAL for training; includes setup, training loop, and evaluation steps.

3 mins read Listed 14 days ago in reddit/mlops/Composite model(Resnet50+FPN+YOLO26 detection head)

huggingface.co Tools and repositories

Polymanius/skyPeak-11n · Hugging Face

This Hugging Face listing provides a YOLO11n model fine-tuned on VisDrone2019-DET to detect people and vehicles in drone images. It reports validation scores, gives a usage example, and notes that its licenses restrict commercial use.

1 min read Listed 9 hours ago in reddit/computervision/I finetuned yolo-11n on a drone dataset and created skypeak-11n. Here are the results!

huggingface.co Tutorial

qihoo360/RevealLayer · Hugging Face

RevealLayer aims to separate visible and hidden RGBA layers from a single RGB image to handle occlusions, supported by a large 100K-image training set and a 200-image benchmark; code and datasets are released for research under Apache 2.0.

2 mins read Listed 7 days ago in reddit/StableDiffusion/TBG ComfyUi Node for Layered Decomposition – RevealLayer

github.com Documentation

GitHub - vyskocj/ImageCLEFdrawnUI2021: Original source code used for ImageCLEF DrawnUI 2021 challenge in: 'Improving web user interface element detection using Faster R-CNN' · GitHub

Installation and configuration guide for ImageCLEF DrawnUI 2021 challenges. It covers required libraries, data layout for screenshot and wireframe tasks, run.py configuration, training commands, model options, greyscale support, and output organization.

4 mins read Listed 28 days ago in reddit/computervision/Looking for ImageCLEF 2021 DrawnUI Wireframe Dataset

tencent.github.io Tools and repositories

WeVisOne-VL: Unified Visual Perception from Objects to Pixels

WeVisOne-VL is a unified vision-language model that combines object- and pixel-level perception using structured next-token prediction, enabling tasks such as detection, localization, counting, pointing, OCR, segmentation, and matting at varying granularities.

1 min read Listed 15 days ago in reddit/computervision/Vision artifacts this week: pixel grounding, long-document OCR, streaming state and robot planning

pyimagesearch.com Tutorial

Build a Real-Time Video Analytics Pipeline with YOLO26 and ByteTrack - PyImageSearch

Build a real-time video analytics pipeline using YOLO26 for detection and ByteTrack for tracking, with line-crossing counting and a live visualization. The system treats detection, tracking, and analytics as separate stages, enabling persistent identities and time-based events.

27 mins read Listed 12 days ago in RSS: PyImageSearch