[*Challenge] image label area select

[yyyanbj]

Prompt[en]

Draw a tight bounding box around the concept stated in the text below. If there is more than one,choose freely.

New types of challenge

New approach (for ex. image watermark)

Sitekey

https://accounts.hcaptcha.com/demo?sitekey=ace50dd0-0d68-44ff-931a-63b670c7eed7

Sitelink

https://accounts.hcaptcha.com/demo?sitekey=ace50dd0-0d68-44ff-931a-63b670c7eed7

Screenshot of the challenge

[QIN2DIM]

https://user-images.githubusercontent.com/62018067/179223582-6a3d7207-d6b8-40f9-8b36-15a30ab3e904.mp4

[QIN2DIM]

I solved it gracefully. Interesting experience.

If you know which website this sitekey belongs to, you can leave a comment here/👻

[M-Zubair10]

This can be done using computer vsion cv2 easily right? by removing background -> finding contours and getBoundingRect()

[QIN2DIM]

😂？？

[QIN2DIM]

So how do you know which object you need to get?

[M-Zubair10]

There is always one object, You said you solved it, how? which model? can I get the code?

[QIN2DIM]

This still doesn't stop the bots from moving, because all the challenges for developers so far are still "mental work" rather than "physical work".

[M-Zubair10]

Did you solved it? I need it

[device-id]

Did you solved it?

I need it

Steps: Predict objects in the image Extract coordinate info (I'm pretty sure the bbox coords are in some pandas output if your use torch hub to predict) Then just click and drag from one coordinate to the next

[QIN2DIM]

It's not that simple 😂

[Pocker88]

Hello Admin, Sir can you add this pack pictures..?!?

[QIN2DIM]

@Pocker88 sure, But I don't have enough data sets right now. Can you provide a stable sitekey for the forest?

[Pocker88]

Sir can you help, how can I make it?!?