add file

.gradio/cached_examples/13/indices.csv

@@ -0,0 +1 @@
+3

.gradio/cached_examples/13/log.csv

@@ -0,0 +1,120 @@
+Image Output,Parsed screen elements,timestamp
+"{""path"": "".gradio/cached_examples/13/Image Output/95caf8d5a60bd2528198/image.webp"", ""url"": ""/gradio_api/file=/tmp/gradio/44a88abb1387b1884333fe3086277174feca52b852454c46389bd04589700217/image.webp"", ""size"": null, ""orig_name"": ""image.webp"", ""mime_type"": null, ""is_stream"": false, ""meta"": {""_type"": ""gradio.FileData""}}","Text Box ID 0: Task Manager
+Text Box ID 1: storage
+Text Box ID 2: InPrivate
+Text Box ID 3: Google
+Text Box ID 4: https:/ WWW.googlecom
+Text Box ID 5: Processes
+Text Box ID 6: Run new task
+Text Box ID 7: End task
+Text Box ID 8: New folder
+Text Box ID 9: Google
+Text Box ID 10: finetune/ldm-ft__
+Text Box ID 11: Gradio
+Text Box ID 12: Pipelines
+Text Box ID 13: Recent
+Text Box ID 14: haotian-liu/LLaVA:
+Text Box ID 15: Processes
+Text Box ID 16: 67%
+Text Box ID 17: 5496
+Text Box ID 18: Status
+Text Box ID 19: CPU
+Text Box ID 20: Memory
+Text Box ID 21: Disk
+Text Box ID 22: About
+Text Box ID 23: Store
+Text Box ID 24: Gmail
+Text Box ID 25: Images
+Text Box ID 26: Sign in
+Text Box ID 27: Performance
+Text Box ID 28: Microscft
+Text Box ID 29: 15.686
+Text Box ID 30: 1,829,9 MB
+Text Box ID 31: MBIs
+Text Box ID 32: Microsoft Team;
+Text Box ID 33: 142.9MB
+Text Box ID 34: MBIs
+Text Box ID 35: App history
+Text Box ID 36: Microsoft Azure Storage Explo_
+Text Box ID 37: Efficiency_
+Text Box ID 38: 0.886
+Text Box ID 39: 245,0 MB
+Text Box ID 40: MBYs
+Text Box ID 41: Startup apps
+Text Box ID 42: WebViewz Manager
+Text Box ID 43: 75,9MB
+Text Box ID 44: MBYs
+Text Box ID 45: Users
+Text Box ID 46: Service Host: Storage Service
+Text Box ID 47: 096
+Text Box ID 48: 1,1 MB
+Text Box ID 49: MBYs
+Text Box ID 50: Details
+Text Box ID 51: Services
+Text Box ID 52: Google
+Text Box ID 53: Google Search
+Text Box ID 54: Feeling Lucky
+Text Box ID 55: Discover
+Text Box ID 56: the ways Chrome keeps you safe while you browse
+Text Box ID 57: Our third decade of climate action: join us
+Text Box ID 58: Settings
+Text Box ID 59: Advertising
+Text Box ID 60: Business
+Text Box ID 61: How Search works
+Text Box ID 62: Privacy
+Text Box ID 63: Terms
+Text Box ID 64: Settings
+Text Box ID 65: 3.53 PM
+Text Box ID 66: Search
+Text Box ID 67: Microsoft
+Text Box ID 68: 10/25/2024
+Text Box ID 69: Edge
+Icon Box ID 70: Microsoft Edge browser.
+Icon Box ID 71: Microsoft 365.
+Icon Box ID 72: Image
+Icon Box ID 73: Image
+Icon Box ID 74: Microsoft Edge browser.
+Icon Box ID 75: Microsoft Edge browser.
+Icon Box ID 76: Teams.
+Icon Box ID 77: Uncomm&ent Selection
+Icon Box ID 78: Microsoft OneNote.
+Icon Box ID 79: Find
+Icon Box ID 80: Microsoft Outlook.
+Icon Box ID 81: Image
+Icon Box ID 82: Maximize
+Icon Box ID 83: Close
+Icon Box ID 84: Dictate
+Icon Box ID 85: Line Spacing
+Icon Box ID 86: Five-point star
+Icon Box ID 87: a search function.
+Icon Box ID 88: Increase
+Icon Box ID 89: More options
+Icon Box ID 90: the Windows operating system.
+Icon Box ID 91: Hyperlink
+Icon Box ID 92: App launcher or menu.
+Icon Box ID 93: Health monitoring
+Icon Box ID 94: Microsoft Outlook.
+Icon Box ID 95: minimizing a window.
+Icon Box ID 96: uBlock Origin (Ctrl+T)
+Icon Box ID 97: Undo
+Icon Box ID 98: Pentagon
+Icon Box ID 99: Settings
+Icon Box ID 100: 1.0%
+Icon Box ID 101: Back
+Icon Box ID 102: Rectangle
+Icon Box ID 103: Redo
+Icon Box ID 104: opening a folder.
+Icon Box ID 105: Justified
+Icon Box ID 106: Label
+Icon Box ID 107: Maximize window
+Icon Box ID 108: Close
+Icon Box ID 109: Close
+Icon Box ID 110: Google Chrome web browser.
+Icon Box ID 111: a loading or progress bar.
+Icon Box ID 112: M0,0L9,0 4.5,5z
+Icon Box ID 113: More options
+Icon Box ID 114: a loading or progress bar.
+Icon Box ID 115: Minimize
+Icon Box ID 116: Undo
+Icon Box ID 117: 0%
+Icon Box ID 118: Draw Functions",2024-10-31 20:12:10.050642

app.py

@@ -12,6 +12,8 @@ from utils import check_ocr_box, get_yolo_model, get_caption_model_processor, ge
 import torch
 from PIL import Image
 
+# Model source: https://huggingface.co/microsoft/OmniParser
+# gr.load("models/microsoft/OmniParser").launch()
 yolo_model = get_yolo_model(model_path='weights/icon_detect/best.pt')
 caption_model_processor = get_caption_model_processor(model_name="florence2", model_name_or_path="weights/icon_caption_florence")
 platform = 'pc'
@@ -58,8 +60,8 @@ DEVICE = torch.device('cuda')
 @spaces.GPU
 def process(
     image_input,
-    box_threshold,
-    iou_threshold
+    box_threshold=0.01,
+    iou_threshold=0.01
 ) -> Optional[Image.Image]:
 
     image_save_path = 'imgs/saved_image_demo.png'