!2294 modify centerface eval_all process

Merge pull request !2294 from zhanghuiyao/modify_centerface

official/cv/centerface/README.md

@@ -194,7 +194,9 @@ step6: eval
 # python setup.py install;
 # cd -; #cd ../../scripts;
 
-bash eval_all.sh [ground_truth_path]
+bash eval_all.sh [GROUND_TRUTH_PATH]
+or
+bash eval_all.sh [GROUND_TRUTH_PATH] [FILTER_EASY](optional) [FILTER_MEDIUM](optional) [FILTER_HARD](optional) [FILTER_SUM](optional)
 ```
 
 - Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
@@ -651,7 +653,9 @@ cd ../../../scripts;
     ```python
     # you need to change the parameter in eval_all.sh
     # default eval the ckpt saved in ./scripts/output/centerface/[89-140]
-    bash eval_all.sh [ground_truth_path]
+    bash eval_all.sh [GROUND_TRUTH_PATH]
+    or
+    bash eval_all.sh [GROUND_TRUTH_PATH] [FILTER_EASY](optional) [FILTER_MEDIUM](optional) [FILTER_HARD](optional) [FILTER_SUM](optional)
     ```
 
 3. test+eval

official/cv/centerface/scripts/eval_all.sh

@@ -14,6 +14,26 @@
 # limitations under the License.
 # ============================================================================
 
+if [ $# != 1 ] && [ $# != 5 ]
+then
+  echo "Usage: bash eval_all.sh [GROUND_TRUTH_PATH]"
+  echo "       bash eval_all.sh [GROUND_TRUTH_PATH] [FILTER_EASY](optional) [FILTER_MEDIUM](optional) [FILTER_HARD](optional) [FILTER_SUM](optional)"
+  exit 1
+fi
+
+FILTER_EASY=0
+FILTER_MEDIUM=0
+FILTER_HARD=0
+FILTER_SUM=0
+
+if [ $# = 5 ]
+then
+  FILTER_EASY=$2
+  FILTER_MEDIUM=$3
+  FILTER_HARD=$4
+  FILTER_SUM=$5
+fi
+
 root=$PWD
 save_path=$root/output/centerface/
 if [ ! -d $save_path ]
@@ -44,9 +64,8 @@ do
 done
 wait
 
-line_number=`awk 'BEGIN{hard=0;nr=0}{if($0~"Hard"){if($4>hard){hard=$4;nr=NR}}}END{print nr}' log_eval_all.txt`
-start_line_number=`expr $line_number - 3`
-end_line_number=`expr $line_number + 1`
-
-echo "The best result " >> log_eval_all.txt
-sed -n "$start_line_number, $end_line_number p" log_eval_all.txt >> log_eval_all.txt
+python ../src/find_best_checkpoint.py --result_file_path=./log_eval_all.txt \
+--filter_easy=$FILTER_EASY \
+--filter_medium=$FILTER_MEDIUM \
+--filter_hard=$FILTER_HARD \
+--filter_sum=$FILTER_SUM >> log_eval_all.txt 2>&1 &

official/cv/centerface/src/find_best_checkpoint.py

+# Copyright 2022 Huawei Technologies Co., Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ============================================================================
+"""find best checkpoint"""
+
+import os
+import argparse
+
+def find_ckpt(file_path, arg):
+    with open(file_path) as f:
+        str_result_list = f.readlines()
+
+    easy_list, medium_list, hard_list, path_list = [], [], [], []
+    easy_index, medium_index, hard_index, sum_index = -1, -1, -1, -1
+    easy_max, medium_max, hard_max, sum_max = -1, -1, -1, -1
+    index = -1
+    for i in range(len(str_result_list)):
+        if str_result_list[i].startswith("==================== Results"):
+            index += 1
+            path_list.append(str_result_list[i][:-1])
+            easy_ap = float(str_result_list[i + 1][15:-1])
+            medium_ap = float(str_result_list[i + 2][15:-1])
+            hard_ap = float(str_result_list[i + 3][15:-1])
+            sum_ap = easy_ap + medium_ap + hard_ap
+            easy_list.append(easy_ap)
+            medium_list.append(medium_ap)
+            hard_list.append(hard_ap)
+            if easy_ap >= arg.filter_easy \
+                    and medium_ap >= arg.filter_medium \
+                    and hard_ap >= arg.filter_hard \
+                    and sum_ap >= arg.filter_sum:
+                if easy_ap >= easy_max:
+                    easy_max = easy_ap
+                    easy_index = index
+                if medium_ap >= medium_max:
+                    medium_max = medium_ap
+                    medium_index = index
+                if hard_ap >= hard_max:
+                    hard_max = hard_ap
+                    hard_index = index
+                if sum_ap >= sum_max:
+                    sum_max = sum_ap
+                    sum_index = index
+
+    if easy_index == -1:
+        print("\nCannot find a checkpoint that meets the filter requirements.")
+    else:
+        print("\nThe best easy result:", flush=True)
+        print(path_list[easy_index], flush=True)
+        print("Easy   Val AP: {}".format(easy_list[easy_index]), flush=True)
+        print("Medium Val AP: {}".format(medium_list[easy_index]), flush=True)
+        print("Hard   Val AP: {}".format(hard_list[easy_index]), flush=True)
+        print("=================================================", flush=True)
+
+        print("\nThe best medium result:", flush=True)
+        print(path_list[medium_index], flush=True)
+        print("Easy   Val AP: {}".format(easy_list[medium_index]), flush=True)
+        print("Medium Val AP: {}".format(medium_list[medium_index]), flush=True)
+        print("Hard   Val AP: {}".format(hard_list[medium_index]), flush=True)
+        print("=================================================", flush=True)
+
+        print("\nThe best hard result:", flush=True)
+        print(path_list[hard_index], flush=True)
+        print("Easy   Val AP: {}".format(easy_list[hard_index]), flush=True)
+        print("Medium Val AP: {}".format(medium_list[hard_index]), flush=True)
+        print("Hard   Val AP: {}".format(hard_list[hard_index]), flush=True)
+        print("=================================================", flush=True)
+
+        print("\nThe best sum result:", flush=True)
+        print(path_list[sum_index], flush=True)
+        print("Easy   Val AP: {}".format(easy_list[sum_index]), flush=True)
+        print("Medium Val AP: {}".format(medium_list[sum_index]), flush=True)
+        print("Hard   Val AP: {}".format(hard_list[sum_index]), flush=True)
+        print("=================================================", flush=True)
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--result_file_path', default='', help='result file')
+    parser.add_argument('--filter_easy', default=0.0, type=float, help='filter easy')
+    parser.add_argument('--filter_medium', default=0.0, type=float, help='filter medium')
+    parser.add_argument('--filter_hard', default=0.0, type=float, help='filter hard')
+    parser.add_argument('--filter_sum', default=0.0, type=float, help='filter sum')
+    args = parser.parse_args()
+
+    if os.path.isfile(args.result_file_path):
+        find_ckpt(args.result_file_path, args)
+    else:
+        raise FileNotFoundError("{} not found.".format(args.result_file_path))