shreyask commited on
Commit
42e1de5
Β·
verified Β·
1 Parent(s): ad0cb9c

readme: clickable model badge, sharper title + gradient

Browse files
Files changed (1) hide show
  1. README.md +18 -7
README.md CHANGED
@@ -1,19 +1,30 @@
1
  ---
2
- title: Falcon Perception
3
  emoji: πŸ¦…
4
- colorFrom: blue
5
- colorTo: indigo
6
  sdk: static
7
  pinned: false
8
  license: apache-2.0
9
- short_description: Open-vocabulary detection in-browser via WebGPU
10
  models:
11
  - tiiuae/Falcon-Perception
12
  - onnx-community/falcon-perception-onnx-webgpu
13
  ---
14
 
15
- # Falcon Perception
16
 
17
- Browser demo for [tiiuae/Falcon-Perception](https://huggingface.co/tiiuae/Falcon-Perception). Image / webcam / video input with Detection / Segment / Tracker render modes. Pixel-accurate segmentation via AnyUp + segm_head, multi-threaded WASM via coi-serviceworker.
18
 
19
- Weights: [onnx-community/falcon-perception-onnx-webgpu](https://huggingface.co/onnx-community/falcon-perception-onnx-webgpu).
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ title: Falcon-Perception-0.6B WebGPU
3
  emoji: πŸ¦…
4
+ colorFrom: indigo
5
+ colorTo: pink
6
  sdk: static
7
  pinned: false
8
  license: apache-2.0
9
+ short_description: Open-vocab detection + segmentation, all in the browser
10
  models:
11
  - tiiuae/Falcon-Perception
12
  - onnx-community/falcon-perception-onnx-webgpu
13
  ---
14
 
15
+ # πŸ¦… Falcon-Perception-0.6B WebGPU
16
 
17
+ A browser demo for **[tiiuae/Falcon-Perception](https://huggingface.co/tiiuae/Falcon-Perception)** β€” a 0.6B open-vocabulary VLM that turns natural-language queries into bounding boxes and pixel-accurate segmentation masks, running fully client-side via WebGPU + ONNX Runtime Web.
18
 
19
+ [![Model](https://img.shields.io/badge/πŸ€—%20Model-tiiuae%2FFalcon--Perception-yellow)](https://huggingface.co/tiiuae/Falcon-Perception)
20
+ [![Weights](https://img.shields.io/badge/πŸ€—%20ONNX%20Weights-onnx--community%2Ffalcon--perception--onnx--webgpu-blue)](https://huggingface.co/onnx-community/falcon-perception-onnx-webgpu)
21
+
22
+ ## What's inside
23
+
24
+ - **Detection** β€” draw bounding boxes for any natural-language query ("athletes", "the runner in front", "mangoes").
25
+ - **Segmentation** β€” pixel-accurate masks via the AnyUp upsampler, all in-browser.
26
+ - **Tracker (preview)** β€” HUD-style reticles on video. Limited by VLM latency between detections; see the in-space disclaimer.
27
+
28
+ ## How it runs
29
+
30
+ 2.4 GB of ONNX weights are fetched once on first visit, then cached by your browser β€” no backend, no API keys, no network round-trip after load. Multi-threaded WASM is enabled via `coi-serviceworker`.