The Segment Anything demo used embeddings we computed ahead of time. This one doesn't: Meta's EfficientSAM — a 2023 FAIR distillation of SAM with a ViT-Tiny encoder — reads your photo and answers your clicks inside this tab, via WebAssembly. nothing is uploaded
Private by construction: the photo is read by your browser and never sent anywhere — you can switch off Wi-Fi after this page loads and it still works. Big photos are shrunk to 1024 px on the long side before the model sees them.
Tap or click the thing you want; add up to six points to grow the selection, or drag a box around it · move the mouse to preview before clicking. Then cut it out and save the sticker. (EfficientSAM has no "remove" points — clear and try again if it grabs too much.)
EfficientSAM-Ti replaces SAM's 636-million-parameter ViT-H image encoder with a ViT-Tiny — small enough to run in WebAssembly on a phone — while reusing SAM's own prompt decoder design.
"SAM-leveraged masked image pretraining": the small encoder is trained to reconstruct the features of SAM's big encoder from masked-out image patches, then fine-tuned on SA-1B. Paper: Xiong et al., CVPR 2024.
The official ONNX encoder (24.8 MB) and decoder (16.6 MB) from Meta's repo, 8-bit quantized on a Raspberry Pi to 6.6 MB + 4.9 MB, served compressed — plus the ~3 MB WebAssembly runtime. Masks agree with the full-precision model at 0.999 IoU on our test photo.
Code and weights are open source. It sits in the Segment Anything family on the Universe map, alongside SAM 2, SAM 3 and SAM 3D.