| license: apache-2.0 | |
| library_name: onnxruntime-web | |
| pipeline_tag: image-segmentation | |
| tags: | |
| - background-removal | |
| - isnet | |
| - onnx | |
| - webgpu | |
| - wasm | |
| - browser | |
| # ISNet (int8) for browser background removal | |
| ISNet segmentation model quantized to int8 (44 MB), exported to ONNX and split | |
| into three parts (`isnet-w8.part00–02`) so every file stays under GitHub's size | |
| limit. Reassemble in the browser with `new Blob(parts)` and load with | |
| [ONNX Runtime Web](https://onnxruntime.ai/docs/tutorials/web/), WASM or WebGPU. | |
| Input 1024×1024 RGB, output single-channel alpha matte. | |
| ## Where it runs | |
| - Live demo (WASM vs WebGPU benchmark): [Space](https://huggingface.co/spaces/chenjindu/browser-remove-background) | |
| - Production tool, free, no upload: **[https://is.tools/remove-background](https://is.tools/remove-background)** | |
| - Source: [github.com/chenjindu/browser-remove-background](https://github.com/chenjindu/browser-remove-background) | |
| Part of [**is.tools**](https://is.tools) — browser-side PDF, image, text and | |
| encoding tools where the file never leaves your device. | |
| ## License | |
| Apache-2.0 (model weights and code). | |