SAM 3 segmentation experiments

Prompt-based image-segmentation experiments in Meta’s SAM 3 notebook, with local environment and prompt changes.

SAM 3PyTorchImage segmentationCUDAJupyter

Study / experimentRank 04 of 4 in this field ↗

Why it sits here. Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.

02 / Closer to the silicon. Closer to the world.STUDY IN SPACE

01 / IMPLEMENTATION & CONTRIBUTION

What the work involves

A small hands-on study modifying example prompts and Python compatibility settings in Meta’s supplied SAM 3 notebook. The model and predictor are upstream implementations.

Technical depth

Text and geometric prompts, confidence thresholds, image-mask inspection, and CUDA mixed-precision inference in the supplied notebook.

The project family

sam3

02 / RESULTS

What came out of it

A locally adapted notebook for inspecting prompt-based segmentation, without a new training or benchmark result.

03 / SUPPORTING EVIDENCE

Follow the source

Implementation notes, project records, and supporting artifacts.

Source context & project scope

This is an upstream-model experiment, not original SAM 3 development. Changes are currently uncommitted and no independent performance evaluation was inspected.

sam3_image_predictor_example.ipynb

Supplied predictor notebook with a locally changed text prompt and executed artifacts.

SOURCE · 2026-09-17

pyproject.toml

Local Python compatibility adjustment against the upstream package.