update nb for resized images (#27)

* update nb for resized images * update checksum for resized data * update random_classifier for resized data * clean object_detector * clean nb
ramp-kits · Dec 6, 2021 · da3ddac · da3ddac
1 parent a7810cb
commit da3ddac
Show file tree

Hide file tree

Showing 3 changed files with 6 additions and 24 deletions.
diff --git a/download_data.py b/download_data.py
@@ -23,7 +23,7 @@
         archive_name="public.tar.gz",
         # to findout checksum use function
         # defined below: hash_folder(folder_path)
-        data_checksum=2471299081,
+        data_checksum=1969228298,
     ),
     "private": dict(
         code="vw8sh", archive_name="private.tar.gz", data_checksum=756276259

diff --git a/follicles_detection_starting_kit.ipynb b/follicles_detection_starting_kit.ipynb
diff --git a/submissions/random_classifier/object_detector.py b/submissions/random_classifier/object_detector.py
@@ -1,22 +1,4 @@
-"""
-Result of `ramp-test --submision random_classifier`
-
-total runtime ~30min
-
-----------------------------
-Mean CV scores
-----------------------------
-score AP <Primordial>    AP <Primary>  AP <Secondary>   AP <Tertiary>         mean AP           time
-train  0.001 ± 0.0007  0.023 ± 0.0077   0.375 ± 0.025  0.436 ± 0.0563  0.209 ± 0.0208  130.8 ± 20.49
-valid       0.0 ± 0.0  0.037 ± 0.0214  0.332 ± 0.0734  0.458 ± 0.0557  0.207 ± 0.0107   166.2 ± 5.91
-test   0.001 ± 0.0018  0.008 ± 0.0141  0.325 ± 0.1782  0.398 ± 0.1259  0.183 ± 0.0683    32.3 ± 0.48
-----------------------------
-Bagged scores
-----------------------------
-score  AP <Primordial>  AP <Primary>  AP <Secondary>  AP <Tertiary>  mean AP
-valid            0.000         0.026           0.361          0.626    0.253
-test             0.002         0.008           0.475          0.625    0.278
-"""
+
 import os
 import numpy as np
 import tensorflow as tf
@@ -128,7 +110,7 @@ def fit(self, X_image_paths, y_true_locations):
         if os.environ.get("RAMP_TEST_MODE", False):
             self._model.fit(X_for_classifier, y_for_classifier, epochs=10)
         else:
-            self._model.fit(X_for_classifier, y_for_classifier, epochs=100)
+            self._model.fit(X_for_classifier, y_for_classifier, epochs=50)
         return self
 
     def predict(self, X):
@@ -145,11 +127,11 @@ def predict(self, X):
     def predict_single_image(self, image_path):
         image = imread(image_path)
 
-        boxes_sizes = [3000, 1000, 300]  # px
+        boxes_sizes = [1500, 1000, 500]  # px
         if os.environ.get("RAMP_TEST_MODE", False):
             boxes_amount = [10, 10, 10]
         else:
-            boxes_amount = [200, 500, 200]
+            boxes_amount = [300, 400, 200]
         boxes = generate_random_windows_for_image(image, boxes_sizes, boxes_amount)
         cropped_images = build_cropped_images(
             image, boxes, crop_size=self.IMG_SHAPE[0:2]