And now with a vectorization step added in:
• SegmentAnything2 (running in the browser with WebGPU to pre-process embeddings) to segment the object at the mouse pointer.
• OpenCV.js to generate a smoothed vector polygon from SAM2 mask.
• Result displayed as SVG.
• SegmentAnything2 (running in the browser with WebGPU to pre-process embeddings) to segment the object at the mouse pointer.
• OpenCV.js to generate a smoothed vector polygon from SAM2 mask.
• Result displayed as SVG.
February 20, 2025 at 8:57 AM
And now with a vectorization step added in:
• SegmentAnything2 (running in the browser with WebGPU to pre-process embeddings) to segment the object at the mouse pointer.
• OpenCV.js to generate a smoothed vector polygon from SAM2 mask.
• Result displayed as SVG.
• SegmentAnything2 (running in the browser with WebGPU to pre-process embeddings) to segment the object at the mouse pointer.
• OpenCV.js to generate a smoothed vector polygon from SAM2 mask.
• Result displayed as SVG.
Wow - this may be old news to some. But TIL that #SegmentAnything2 runs pretty well *entirely* in the browser. Not just the segmentation–also generating the embeddings 🤯 medium.com/@geronimo7/i...
Image Segmentation in the Browser with Segment Anything Model 2
Run Meta’s SAM2 with onnxruntime-web
medium.com
February 19, 2025 at 12:13 PM
Wow - this may be old news to some. But TIL that #SegmentAnything2 runs pretty well *entirely* in the browser. Not just the segmentation–also generating the embeddings 🤯 medium.com/@geronimo7/i...
Dans le cadre du programme #ISIDORE 2030, l'équipe du Huma-Num Lab va participer au colloque Humanistica 2025 (Dakar) avec une communication d'un des futurs outils en cours de développement pour @isidore-science.bsky.social : “SegmentArt” hnlab.huma-num.fr/blog/2025/04... #IA
SegmentArt, une méthode plus rapide pour annoter des images en utilisant SegmentAnything2
Présentation de l'intervention du colloque Humanistica 2025 (Dakar) proposé par le HN Lab sur les questions de l'annotation d'images (SegmentArt) par l'utilisation SegmentAnything2
hnlab.huma-num.fr
April 15, 2025 at 5:51 PM
Dans le cadre du programme #ISIDORE 2030, l'équipe du Huma-Num Lab va participer au colloque Humanistica 2025 (Dakar) avec une communication d'un des futurs outils en cours de développement pour @isidore-science.bsky.social : “SegmentArt” hnlab.huma-num.fr/blog/2025/04... #IA
It uses SegmentAnything2. The "tiny" model runs pretty well in the browser, if WebGPU is available (only Chrome and Edge for now).
February 27, 2025 at 8:23 AM
It uses SegmentAnything2. The "tiny" model runs pretty well in the browser, if WebGPU is available (only Chrome and Edge for now).
A zero‑shot framework using Segment Anything 2 tracks cells in 2‑D and large‑scale 3‑D time‑lapse microscopy without any dataset‑specific training; the preprint was submitted on 12 Sep 2025. https://getnews.me/zero-shot-cell-tracking-using-segment-anything-2-ai-model/ #segmentanything2 #celltracking
September 17, 2025 at 1:27 AM
A zero‑shot framework using Segment Anything 2 tracks cells in 2‑D and large‑scale 3‑D time‑lapse microscopy without any dataset‑specific training; the preprint was submitted on 12 Sep 2025. https://getnews.me/zero-shot-cell-tracking-using-segment-anything-2-ai-model/ #segmentanything2 #celltracking