Alternatives to Potato

CVAT is an open-source data annotation platform for image, video, 3D point cloud, and audio labeling.

Open-source Python framework for generating training datasets from sandboxed tool execution traces

Open-source multimodal data engine that turns raw video, images, and audio into training-ready AI datasets.

Velvet sources, verifies, and delivers curated video datasets so AI labs can train interactive and world models with spatial reasoning.

A free 70,000-image Zalando clothing dataset built as a drop-in replacement for MNIST.

Offline AI image annotation with local SAM2, SAM3 and YOLO-World — one-time purchase from $49.

Human-recorded computer-use datasets that teach AI agents to operate real software the way people do.

AfterQuery captures expert reasoning and structures it into SFT, RL rubric, agent, and computer-use training data for frontier models.

Cortex AI supplies real-workplace egocentric video and robot trajectory data for training embodied AI and robot foundation models.

Snorkel AI builds the expert-curated datasets, evaluation rubrics, and runnable environments that fix where frontier models and agents break.

Browser and offline desktop data labeling for image, text, document, and video annotation.

BioStack Platforms supplies structured clinical datasets and RL environments for healthcare AI teams.

Open-source data labeling platform for AI evaluation, computer vision, document AI, and human-in-the-loop review across every modality.

Raiinmaker sources rights-cleared custom video datasets and runs real-time human evaluation for generative AI video models.

Early-stage San Francisco robotics lab building policies for robots that do real work, and hiring its first researchers.

Applied research lab shipping autoscaled RL data and self-evolving benchmarks for tool-use and computer-use agents.

GameEngineBench scores frontier AI models on real game-engine programming — see which model can actually ship UE5 code.

Prolific connects AI and research teams to 300,000+ verified participants for human feedback, preference tuning, and safety evaluations.

Mundo AI builds culturally-aware audio and video training data, evaluations, and applied research for teams working on perceptual AI.

Spotlight is Renumics' open-source Python library for visualizing and curating unstructured datasets as interactive similarity maps.

Markup turns PDFs and long text into structured NLP and ML datasets with AI-assisted annotation in the browser

Open-source Python library to load, stream, and preprocess Hugging Face datasets for AI model training

PublicAI pays a decentralized workforce in crypto for multi-modal AI training data, annotation, and real coding trajectories.

LanceDB is an open-source multimodal lakehouse that keeps curation, feature engineering, hybrid search, and training on one Lance table.

Labelbox builds RL environments, expert-labeled training data, and Recursion agents that grade their own outcomes

Agent training data, RL environments, and safety red-teaming built by Toloka for LLM and agent teams

Autodistill auto-labels image datasets with foundation models, then trains small vision models you own.

Structured3D is a free 3.5K-design synthetic dataset for 3D structure and room layout research.

Private-deployment data annotation platform for multimodal AI training, from BasicAI.

SWE-smith generates training data for software engineering agents from any GitHub repository.