loading tag register... ok
loading matching posts... ok
loading related topics... ok
teds read --tag vision-multimodal
Vision & Multimodal
Hands-on work with computer vision, vision-language models, object detection, visual grounding, and multimodal systems.
Related service
Need help with a vision or multimodal AI system? Talk to us about AI engineering consulting.
Construction-Site-Traversability: Robot Loops With Camera, Range and Lidar in FiftyOne
A tracked robot drives closed loops through two active construction sites. Scrub RGB, range, lidar, GNSS and odometry together in FiftyOne.
BlogGR00T-X-Embodiment-Sim: Seven Robot Bodies, 42 Episodes, Compared in FiftyOne
GR00T-X-Embodiment-Sim has six episodes from each of seven robot bodies. Compare cameras and joint telemetry on one timeline in FiftyOne.
Blogegodex-test-multimodal: 68-Joint Hand and Body Pose With Egocentric Video in FiftyOne
EgoDex pairs Apple Vision Pro video with 3D pose for 68 body and finger joints. Browse 3,243 test episodes with a 3D skeleton in FiftyOne.
Blog2026-Humanoid-IKEA-Assembly-Challenge: 35 Motors, Seven Cameras, in FiftyOne
Six humanoid furniture-assembly episodes with 35 motors, seven camera streams and annotated subtasks. Scrub torque plots next to wrist cameras in FiftyOne.
BlogSEW-Multimodal-AMR: RGB, Thermal, 4D Radar and ToF on One Robot, Loaded in FiftyOne
SEW's mobile robot carries RGB, thermal, ToF, 4D radar, laser scanners and ultrasonics. Compare 3,151 labeled frames across weather and light in FiftyOne.
BlogHapTile: 1,699 UR5e Episodes With Tactile Video and Haptic Levels, in FiftyOne
HapTile pairs UR5e teleoperation with two fingertip tactile streams, two RGB-D cameras and haptic levels. Scrub 1,699 episodes in FiftyOne.
BlogWhy Squeaky-Clean Datasets Break Production Vision Models
Over-cleaning training data strips structured real-world noise, leaving vision models brittle. Build a data preservation list before running cleaning sprints.
BlogA 3-Step Data Quality Pre-Flight Check Before Training Vision Models
New architectures drop every 33 hours, but training success depends on data fundamentals. Run this 3-step audit on resolution, labels, and class balance.
BlogScene Complexity Metrics: Measuring Clutter and Density Beyond mAP
Aggregate mean average precision hides operational failure modes. Use visual clutter scores, instance density, and background complexity to stress-test models.
BlogHands-on Point Tracking in Video with Meta AI CoTracker3 and FiftyOne
Meta AI released CoTracker3 for dense point tracking across video frames. Learn how point tracking works, its mechanics, and how to visualize tracks in FiftyOne.
BlogWhy Duplicate Training Data Silently Corrupts Vision Models
Duplicate images inflate mAP, distort precision-recall curves, and create artificial confidence. Here is how to identify clusters and selectively prune.
BlogDetecting Hidden Outliers in Visual AI Datasets with Embeddings and UMAP
Uncover silent data anomalies before production failure. How to use foundation model embeddings, Local Outlier Factor (LOF), and UMAP projections.
BlogBuilding a Data Quality Feedback Loop for Visual AI Systems
Stop wasting engineering sprints tuning hyperparameters on flawed datasets. Here is how to build a 4-step data-centric evaluation loop for computer vision.
BlogCritical Image Quality Metrics That Break Computer Vision in Production
Vision models fail when real-world camera conditions diverge from training data. How to measure and audit blur, exposure, contrast, and compression artifacts.
BlogEgocentric-EMG-Force: First-Person RGB-D With Forearm EMG, in FiftyOne
Egocentric-EMG-Force pairs first-person RGB-D video with EMG from both wrists, hand skeletons and finger force estimates across 8 household tasks.
BlogDreamTac: 703 Franka Trajectories With Gel Fingertip Video for Robot Learning
DreamTac records a Franka arm with a wrist camera, a third-person camera and two gel fingertips. Watch contact in four synced views in FiftyOne.
BlogTII-RATM-Drone-Racing: Measure Odometry Drift Against Motion Capture in FiftyOne
TII-RATM pairs a racing drone's own odometry with motion-capture truth for six flights. Sort by tracking error and watch the feature tracks.
Blog10 Open-Source Robotics Repositories for 3D Perception, VLA Policies, and Sensor Calibration
An inspectable technical breakdown of 10 open-source robotics and physical AI GitHub repositories solving training-free 3D occupancy, targetless calibration, 50Hz VLA latency, and humanoid memory.
BlogBoilingBench-Multimodal: Scrub Boiling Video Against Temperature and Sound
BoilingBench-Multimodal syncs high-speed video, surface temperature and three acoustic sensors on one clock across 7 boiling and cooling runs.
BlogTaF-Tactile-Force: Tactile Images, Pressure Maps and Six-Axis Force on One Clock
TaF-Tactile-Force syncs a tactile camera, a 12x12 pressure array and a six-axis force sensor at 30 fps, with contacts up to 90.8 N.
BlogRoboLab-EgoX: Filter 3,368 Labeled Robot Failures in FiftyOne
RoboLab-EgoX keeps all 4,000 RoboLab policy rollouts, 632 successes and 3,368 failures, with three camera views and depth. Filter them in FiftyOne.
BlogMirrorSentinel-Elevator: Checking Lidar Against Hand-Measured Elevator Cabins
MirrorSentinel-Elevator pairs a ZED2 camera and an Ouster lidar with hand-measured cabin boundaries, so you can see what a mirror does to range.
BlogRobots Can See. They Still Can't Feel: 8 Tactile LeRobot Datasets in FiftyOne
Why touch is the missing frontier for robot learning, how tactile data is structured across gel sensors, force vectors, and taxel grids, and 8 open LeRobot datasets loaded into FiftyOne.
BlogRobot Data Quality Scoring: Why Scores Are Triage, Not Gates
Why action-only smoothness metrics invert on critical structural errors, why detector AUROC does not predict policy success, and how to use automated scoring for triage in FiftyOne.
BlogMCAP Authoring Errors: 6 Silent Bugs and How to Fix Them
When authoring MCAP files from robotics logs, errors rarely throw runtime exceptions. The file writes clean and passes validators, yet fails in viewers. Here are the 6 silent bugs and how to fix them.
BlogFrom Franka Arms to Humanoids: 11 LeRobot v3.0 Datasets for VLA Training in 2026
A comprehensive guide to 11 robot-learning datasets converted into LeRobot v3.0 format for inspection in FiftyOne, covering arms, bimanual setups, humanoids, and the robotics data action gap.
BlogImage Copy Detection: Find Edited Copies with pHash, CLIP, and FiftyOne
Why single-signal duplicate detection fails on cropped, mirrored, or filtered images, and how Image Link Lab pairs pHash and CLIP embeddings under configurable logic in FiftyOne.
BlogEmbeddingGemma 2 in FiftyOne: Search Video by Text, Sound, or Your Webcam
How to use Google DeepMind's EmbeddingGemma 2 inside FiftyOne with LanceDB to search across 10,000 video clips using text, soundtrack audio, or webcam feeds in a shared 768-D space.
BlogData Curation, the Voxel51 Agent, and In-App Annotation: A CropAndWeed Walkthrough
A data-centric audit of the CropAndWeed dataset exposing class imbalance misconceptions, sampling biases, 370 degenerate boxes, and 23 stem-point defects using FiftyOne.
BlogBuilding Production-Ready AI Document Understanding Pipelines with GLM-OCR
Why traditional OCR pipelines break on tables and complex layouts, and how GLM-OCR extracts native markdown, LaTeX formulas, and structured JSON inside FiftyOne.
BlogMultimodal AI Changes the Shape of Software
How text, image, audio, video, and document models change product design by letting users ask, point, upload, inspect, and correct.
BlogTeaching PaliGemma 2 to Spot Construction-Site Hazards
Fine-tune PaliGemma 2 with Transformers 5, QLoRA, and a construction safety dataset so a vision-language model can return object labels and bounding boxes for job-site hazards.
BlogPreference-Aligning Vision Models: DPO Fine-Tuning Qwen3.5-4B in Transformers 5.12.1
A modern rewrite of an older SmolVLM notebook: align Qwen3.5-4B with Direct Preference Optimization (DPO), LoRA adapters, and current Transformers + TRL APIs.
BlogModern Visual Question Answering Fine-Tuning: Qwen3-VL-4B with QLoRA in Transformers 5.12.1
A current replacement for an older PaliGemma notebook: fine-tune Qwen3-VL-4B on a small VQAv2 split with QLoRA, modern Transformers APIs, and a cleaner multimodal training loop.
BlogModerating Memes with Qwen2.5-VL: Zero-Shot Hateful Content Detection in Transformers 5.12.1
A cleaned-up multimodal moderation walkthrough that uses Qwen2.5-VL with Transformers 5.12.1, the Hateful Memes validation split, and a reproducible zero-shot evaluation script.
BlogHuman Preference Tuning for Small VLMs: SmolVLM2 + DPO in Transformers 5.12.1
A modern guide to preference-tuning SmolVLM2 with Direct Preference Optimization, TRL, PEFT LoRA adapters, and the current Transformers image-text API.
BlogFrom Prompt to Pixels: Generate Images with Qwen-Image, Diffusers, and Transformers
A current text-to-image tutorial using Qwen-Image, Hugging Face Diffusers, and the latest Transformers stack, with reproducible prompts, aspect ratios, seeds, and a fast FLUX.1-schnell fallback.
BlogFrom Messy Labels to Production Boxes: Ultralytics YOLO26 + Transformers 5.12.1
A modern, reproducible object detection workflow that replaces legacy SuperGradients/YOLO-NAS setups with Ultralytics YOLO26 and current Transformers tooling.
BlogTrain a Modern VLM on One GPU: Qwen3-VL with Unsloth and Transformers 5.12.1
A current vision-language fine-tuning guide that replaces an old Qwen2-VL notebook with a cleaner Qwen3-VL workflow, modern Transformers APIs, and a practical DocumentVQA example.
BlogFine-Tuning Qwen3.6-VL for Brain Tumor MRI Detection with Transformers 5
A practical, up-to-date tutorial for adapting Qwen3.6-VL to brain tumor MRI yes/no detection with Transformers 5, QLoRA, and reproducible multimodal training.
BlogFine-Tune ViTPose++ for Keypoint Detection with Transformers
A practical Transformers-native guide to fine-tuning ViTPose++ on COCO-style keypoints with generated heatmap targets, RT-DETR inference, and current pose-estimation caveats.
BlogFind What You Mean: Zero-Shot Visual Grounding with Qwen3-VL and Transformers
Use Qwen3-VL with the latest Transformers API to detect objects, ground natural-language phrases, and visualize bounding boxes from structured multimodal outputs.
BlogDETR Explained: Set Prediction Object Detection in Transformers 5.12.1
Learn DETR object detection with a modern Transformers 5.12.1 example, pinned requirements, and a clean explanation of set prediction, Hungarian matching, and object queries.
BlogDetect What You Can Name: PaliGemma 2 Object Detection with Transformers
A practical tutorial for using current PaliGemma 2 mix checkpoints with Transformers 5 to run prompt-driven object detection and draw bounding boxes from generated location tokens.
BlogCaption by Consensus: Ranking Image Descriptions with SigLIP2 and Transformers
Use SigLIP2 with the latest Transformers APIs to score candidate image captions, choose the best description, and understand when a contrastive vision-language encoder is the right tool.
BlogYOLOv12 in Practice: A Real-Time Object Detection Guide
A practical guide to YOLOv12's attention-centric real-time detector, with a current install path, an inference example, and notes on the official detection, segmentation, and classification weights.