{"name":"tamserve","repo":"aryamandarda/tamserve","url":"https://midorreal.com/project/aryamandarda-tamserve","repository":"https://github.com/aryamandarda/tamserve","description":"Live, multi-camera object tracking with SAM3, SAM 2 and EdgeTAM, on one GPU or many. One model per GPU serves every camera: faster than the original code, about 4x less GPU memory, the same accuracy.","facts":{"license":{"value":"Apache-2.0","source":"https://github.com/aryamandarda/tamserve","checked":"21 Nov 2026"},"language":{"value":"Python","source":"https://github.com/aryamandarda/tamserve","checked":"21 Nov 2026"},"last_release":{"value":"v0.3.3, 19 Nov 2026","source":"https://github.com/aryamandarda/tamserve/releases","checked":"21 Nov 2026"},"releases":{"value":"6","source":"https://github.com/aryamandarda/tamserve/releases","checked":"20 Nov 2026"},"contributors":{"value":"1","source":"https://github.com/aryamandarda/tamserve/graphs/contributors","checked":"20 Nov 2026"},"archived":{"value":"no","source":"https://github.com/aryamandarda/tamserve","checked":"21 Nov 2026"}},"stars":269.0,"writeup":{"what_it_is":"tamserve is a serving engine for SAM 2, SAM 2.1 and EdgeTAM object tracking. One copy of the model tracks objects across multiple live camera feeds on one or more GPUs, and it can also find objects by name.","audience":"Developers building live multi-camera video object tracking systems","claims":[{"kind":"specific","claim":"Uses about 4x less GPU memory with 8 cameras","excerpt":"GPU memory, 8 cameras | 8.6 GB | 2.2 GB","status":"not_checked"},{"kind":"specific","claim":"Faster than the original code in the README's benchmarks","excerpt":"Speed, 1 camera | 177 frames/s | 219 frames/s","status":"not_checked"},{"kind":"specific","claim":"Accuracy on DAVIS 2017 is essentially unchanged","excerpt":"Accuracy on DAVIS 2017 | 0.8686 | 0.8685","status":"not_checked"},{"kind":"specific","claim":"Can track objects named by a word or phrase","excerpt":"Give it a phrase, or several, and it finds every match on every camera","status":"not_checked"},{"kind":"specific","claim":"Supports multiple GPUs through a Router","excerpt":"a `Router` places the cameras and gathers every result","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-18"},"why_now":null,"owner_supplied":null,"score":{"verdict":"overhyped","hype":100,"reality":47,"gap":53,"momentum":null,"confidence":85,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T07:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-11-19T03:07:03+00:00","text":"Latest release v0.3.3","source":"https://github.com/aryamandarda/tamserve/releases"},{"date":"2026-11-11T21:52:37+00:00","text":"First release","source":"https://github.com/aryamandarda/tamserve/releases"},{"date":"2026-11-11T08:25:48+00:00","text":"Repository created","source":"https://github.com/aryamandarda/tamserve"}],"cite":"tamserve is a GitHub project at https://github.com/aryamandarda/tamserve written mainly in Python, described there as \"Live, multi-camera object tracking with SAM3, SAM 2 and EdgeTAM, on one GPU or many. One model per GPU serves every camera: faster than the original code, about 4x less GPU memory, the same accuracy\".","attribution":"Data from Mid or Real, https://midorreal.com/project/aryamandarda-tamserve, checked 22 November 2026. Attribution with a link is required.","still_real":null}