trim spacing
Browse files
README.md
CHANGED
|
@@ -46,12 +46,10 @@ from PIL import Image
|
|
| 46 |
from askui_ml_helper.utils.pta_text import PtaTextInference
|
| 47 |
|
| 48 |
pta_text_inference = PtaTextInference("pta-text-v0.1.pt")
|
| 49 |
-
|
| 50 |
url = "https://docs.askui.com/assets/images/how_askui_works_architecture-363bc8be35bd228e884c83d15acd19f7.png"
|
| 51 |
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
|
| 52 |
prompt = 'click on the text "Operating System"'
|
| 53 |
|
| 54 |
-
|
| 55 |
render_image = pta_text_inference.process_image_and_draw_circle(image, prompt, radius=15)
|
| 56 |
render_image.show()
|
| 57 |
>>> Uploaded image with "a red dot", where click operation is predicted
|
|
@@ -67,12 +65,10 @@ from PIL import Image
|
|
| 67 |
from askui_ml_helper.utils.pta_text import PtaTextInference
|
| 68 |
|
| 69 |
pta_text_inference = PtaTextInference("pta-text-v0.1.pt")
|
| 70 |
-
|
| 71 |
url = "https://docs.askui.com/assets/images/how_askui_works_architecture-363bc8be35bd228e884c83d15acd19f7.png"
|
| 72 |
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
|
| 73 |
prompt = 'click on the text "Operating System"'
|
| 74 |
|
| 75 |
-
|
| 76 |
coordinates = pta_text_inference.process_image(image, prompt)
|
| 77 |
coordinates
|
| 78 |
>>> [0.3981265723705292, 0.13768285512924194]
|
|
|
|
| 46 |
from askui_ml_helper.utils.pta_text import PtaTextInference
|
| 47 |
|
| 48 |
pta_text_inference = PtaTextInference("pta-text-v0.1.pt")
|
|
|
|
| 49 |
url = "https://docs.askui.com/assets/images/how_askui_works_architecture-363bc8be35bd228e884c83d15acd19f7.png"
|
| 50 |
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
|
| 51 |
prompt = 'click on the text "Operating System"'
|
| 52 |
|
|
|
|
| 53 |
render_image = pta_text_inference.process_image_and_draw_circle(image, prompt, radius=15)
|
| 54 |
render_image.show()
|
| 55 |
>>> Uploaded image with "a red dot", where click operation is predicted
|
|
|
|
| 65 |
from askui_ml_helper.utils.pta_text import PtaTextInference
|
| 66 |
|
| 67 |
pta_text_inference = PtaTextInference("pta-text-v0.1.pt")
|
|
|
|
| 68 |
url = "https://docs.askui.com/assets/images/how_askui_works_architecture-363bc8be35bd228e884c83d15acd19f7.png"
|
| 69 |
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
|
| 70 |
prompt = 'click on the text "Operating System"'
|
| 71 |
|
|
|
|
| 72 |
coordinates = pta_text_inference.process_image(image, prompt)
|
| 73 |
coordinates
|
| 74 |
>>> [0.3981265723705292, 0.13768285512924194]
|