Fix: rasanc max_samples_by_conf to not return negative (#2897)

* added failing test

* added the bugfix of RANSAC.max_samples_by_conf

* fixing broken links

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

CONTRIBUTING.md

@@ -9,12 +9,12 @@ Everyone is welcome to get involved with the project. There are different ways t
       1. using the GitHub discussion at Kornia repo: [GH Discussions](https://github.com/kornia/kornia/discussions)
       2. our slack workspace to keep in touch with our core contributors and the community:
          [Join Here](https://join.slack.com/t/kornia/shared_invite/zt-csobk21g-2AQRi~X9Uu6PLMuUZdvfjA)
-      3. using the `#kornia` tag in [PyTorch Discuss](https://discuss.pytorch.org>)
+      3. using the `#kornia` tag in [PyTorch Discuss](https://discuss.pytorch.org)
     - Please, don't use GitHub issues for Q&A.
     - In case you are a developer and want to learn more about the PyTorch ecosystem, we suggest you join the PyTorch
-      slack. You can apply using this form: [https://bit.ly/ptslack](https://bit.ly/ptslack>)
+      slack. You can apply using this form: [https://bit.ly/ptslack](https://bit.ly/ptslack)
 
-2. Report bugs through [GitHub issues](https://github.com/kornia/kornia/issues>):
+2. Report bugs through [GitHub issues](https://github.com/kornia/kornia/issues):
    - Do a quick search first to see whether others reported a similar issue.
    - In case you find an unreported bug, please open a new ticket.
    - Try to provide as much information as possible. Report using one of the available templates. Some tips:

kornia/geometry/ransac.py

@@ -103,7 +103,7 @@ def max_samples_by_conf(n_inl: int, num_tc: int, sample_size: int, conf: float)
             return 1.0
         if n_inl == num_tc:
             return 1.0
-        return math.log(1.0 - conf) / max(eps, math.log(max(eps, 1.0 - math.pow(n_inl / num_tc, sample_size))))
+        return math.log(1.0 - conf) / min(-eps, math.log(max(eps, 1.0 - math.pow(n_inl / num_tc, sample_size))))
 
     def estimate_model_from_minsample(self, kp1: Tensor, kp2: Tensor) -> Tensor:
         batch_size, sample_size = kp1.shape[:2]

tests/geometry/test_ransac.py

@@ -1,3 +1,5 @@
+import math
+
 import pytest
 import torch
 
@@ -241,3 +243,17 @@ def gradfun(p1, p2):
             return model(p1, p2)[0]
 
         self.gradcheck(gradfun, (points1, points2), fast_mode=False, requires_grad=(True, False, False))
+
+
+class TestRansacMethods:
+    def test_max_samples_by_conf(self):
+        conf = 0.99
+        eps = 1e-9
+
+        x = RANSAC.max_samples_by_conf(n_inl=1, num_tc=1000, sample_size=7, conf=conf)
+        assert x > 0.0
+        assert x == math.log(1.0 - conf) / -eps
+
+        x = RANSAC.max_samples_by_conf(n_inl=999999999, num_tc=1000000000, sample_size=1, conf=conf)
+        assert x > 0.0
+        assert x == math.log(1.0 - conf) / math.log(eps)