Torchvision DenseNet121 initialized from DenseNet121_Weights.IMAGENET1K_V1, fine-tuned for 4-class image classification. Bounding boxes are not used. This is a research classifier; it is not ...

Torchvision MobileNetV2 initialized from MobileNet_V2_Weights.IMAGENET1K_V2, fine-tuned for 4-class image classification. Bounding boxes are not used. This is a research classifier; it is not ...

DG6 is a LeRobot dataset (format v3.0) for one skill on an SO-101 arm: grip the green box. Each episode starts with the gripper above or near a 32 mm green box and ends after the box is lifted.

This dataset is an adversarial guardrails benchmark for agent services. It is designed to test whether guardrails correctly block harmful or disallowed agent actions while still allowing legitimate, ...

The task datasets for WorkBuddy Bench, a benchmark for evaluating coding agents on real-world developer, PM, algo, QA, ops, and security work. This repository hosts only the task data; the evaluation ...

Synaps-Image is not a new image model. It is tooling that Synaptivo adds on top of a strong open generator, FLUX 3 Image Klein with its prompt upsampler PUP, both by Black Forest Labs. The tooling ...

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Binary semantic segmentation of building footprints from high-resolution orthophotos (aerial / satellite imagery). Trained with MMSegmentation using an HRNet-W48 backbone and an FCN head. This is the ...

Analysis of particle approximations for perturbed Wasserstein gradient flow shows that curvature dynamics govern long-horizon tracking accuracy, with temporary instability offset by recovery.

An int8 ONNX copy of OpenAI's English-only Whisper small model, for on-device speech recognition with ONNX Runtime. The Thì Thầm app uses it for English. It is ...

SACQ is an ultra-low-bit weight-only quantization framework for vision-language-action models that uses structure-aware codebook fitting, action-aware mixed-bit allocation, and post-quantization ...

How to use ctcheung/microduck-sock-heist-6m-progress12-v3-from1500-10k with Microduck: # Replace SLOT with the slot specified in the model card (walk, stand, sitstand, ground_pick, kick_left, ...