I found some mistakes in the way you in unwarp character unwarp the character bounding boxes

[shitvuive123]

I don't think it is the right logic code to match each character box location to real location in the original image. Here is my code: for j in len(bboxes): real_bboxes = [] for pts in bboxes[j]: point = np.append(point, 1) assert len(point) == 3 tmp = np.matmul(inv(MM), point) print('tmp', tmp) real_point = tmp / tmp[-1] real_bboxes.append(real_point) bboxes[j] = np.array(real_bboxes)

[backtime92]

@shitvuive123 thanks for your comment, i will check it.

[Lovegood-1]

Hi,

I did corresponding experiences. The results shows that there is nearly no difference between them,

considering we use int8. But there is something different when you use theoretical analysis.

when I was using a Craft repo to do text detection. Then, I did corresponding experiences.

---

Conclusion 1

The results shows that there is nearly no difference between them,

considering we use int8.

---

Conclusion 2

But there is something different when you use theoretical analysis.

---

Reason 1

And the MM is array([[ 9.94309506e-01, -9.94309526e-02, 3.93746573e+01], [ 7.17308145e-02, 9.90781875e-01, -4.22288271e+02], [-5.43546480e-11, -3.72669030e-10, 1.00000000e+00]]).

the [-5.43546480e-11, -3.72669030e-10, 1.00000000e+00] in MM makes the last dimention unimportant. """