Sigmoid overflow error for very small input values z. and Colorful drawing boxes

.gitignore

@@ -0,0 +1,3 @@
+*.bak
+*.h5
+__pycache__/*

README.md

@@ -11,6 +11,51 @@ This repo contains the implementation of YOLOv2 in Keras with Tensorflow backend
 - [ ] Multiscale training
 - [ ] mAP Evaluation
 
+
+## Kabrau modifications:
+
+### Evaluation 
+
+Validation and test evaluate, by classes, mAP, FPS, and PR CURVE image
+- Important: Clone https://github.com/kabrau/mean_average_precision
+
+`python evaluate.py -c confg.json -w /path/to/best_weights.h5 -i /path/validation_images/ -a /path/validation_pascal_xml -j /path/test_images/ -t /path/test_pascal_xml`
+
+Example
+
+`python evaluate.py -c E:/rodney/configs/keras-yolo2/config.Inception3.json -w E:/rodney/weights/Inception3_V2c.h5 -i E:/rodney/DSDFinal/images/ -a E:/rodney/DSDFinal/split/validation/ -j E:/rodney/DSDFinal/images/ -t E:/rodney/DSDFinal/split/test/`
+
+Result example
+
+```
+Valid Evaluate
+ascending_stair 0.8334
+descending_stair 0.6084
+door 0.8277
+elevator_door 0.9268
+Valid mAP: 0.7991
+
+Total Images:  546
+Elapse Time: 11.0589
+Avg Image Time: 0.0203
+FPS:  49
+
+Test Evaluate
+ascending_stair 0.7365
+descending_stair 0.6934
+door 0.7482
+elevator_door 0.9024
+Test mAP: 0.7701
+
+Total Images:  535
+Elapse Time: 9.8113
+Avg Image Time: 0.0183
+Desvio Padrão: 0.0059
+FPS:  54
+```
+
+
+
 ## Some example applications (click for videos):
 
 ### Raccon detection

evaluate.py

@@ -0,0 +1,112 @@
+#! /usr/bin/env python
+
+import argparse
+import os
+import numpy as np
+from preprocessing import parse_annotation
+from frontend_evaluate import YOLO
+import json
+
+os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"
+os.environ["CUDA_VISIBLE_DEVICES"]="0"
+
+argparser = argparse.ArgumentParser(
+    description='Train and validate YOLO_v2 model on any dataset')
+
+argparser.add_argument(
+    '-c',
+    '--conf',
+    help='path to configuration file')
+
+argparser.add_argument(
+    '-w',
+    '--weights',
+    help='path to pretrained weights')  
+
+argparser.add_argument(
+    '-i',
+    '--inputVal',
+    help='val image folder')
+
+argparser.add_argument(
+    '-a',
+    '--annotVal',
+    help='val annotation folder')    
+
+argparser.add_argument(
+    '-j',
+    '--inputTest',
+    help='test image folder')
+
+argparser.add_argument(
+    '-t',
+    '--annotTest',
+    help='test annotation folder')    
+
+def _main_(args):
+    config_path = args.conf
+
+    with open(config_path) as config_buffer:    
+        config = json.loads(config_buffer.read())
+
+    ###############################
+    #   Parse the annotations 
+    ###############################
+
+    valid_imgs, valid_labels = parse_annotation(args.annotVal, 
+                                                args.inputVal, 
+                                                config['model']['labels'])
+
+    test_imgs, test_labels = parse_annotation(args.annotTest, 
+                                              args.inputTest, 
+                                              config['model']['labels'])
+
+
+    if len(config['model']['labels']) > 0:
+        overlap_labels = set(config['model']['labels']).intersection(set(valid_labels.keys()))
+
+        print('Seen labels:\t', valid_labels)
+        print('Given labels:\t', config['model']['labels'])
+        print('Overlap labels:\t', overlap_labels)           
+
+        if len(overlap_labels) < len(config['model']['labels']):
+            print('Some labels have no annotations! Please revise the list of labels in the config.json file!')
+            return
+    else:
+        print('No labels are provided. Train on all seen labels.')
+        config['model']['labels'] = valid_labels.keys()
+
+    ###############################
+    #   Construct the model 
+    ###############################
+
+    yolo = YOLO(backend             = config['model']['backend'],
+                input_size          = config['model']['input_size'], 
+                labels              = config['model']['labels'], 
+                max_box_per_image   = config['model']['max_box_per_image'],
+                anchors             = config['model']['anchors'])
+
+    ###############################
+    #   Load the pretrained weights (if any) 
+    ###############################    
+
+    print("Loading pre-trained weights in", args.weights)
+    yolo.load_weights(args.weights)
+
+    ###############################
+    #   Start the training process 
+    ###############################
+
+    yolo.eval(valid_imgs         = valid_imgs,
+               test_imgs          = test_imgs,               
+               learning_rate      = config['train']['learning_rate'], 
+               batch_size         = config['train']['batch_size'],
+               object_scale       = config['train']['object_scale'],
+               no_object_scale    = config['train']['no_object_scale'],
+               coord_scale        = config['train']['coord_scale'],
+               class_scale        = config['train']['class_scale'],
+               debug              = config['train']['debug'])
+
+if __name__ == '__main__':
+    args = argparser.parse_args()
+    _main_(args)

frontend.py

@@ -315,7 +315,7 @@ def train(self, train_imgs,     # the list of images to train the model
                                      save_best_only=True, 
                                      mode='min', 
                                      period=1)
-        tensorboard = TensorBoard(log_dir=os.path.expanduser('~/logs/'), 
+        tensorboard = TensorBoard(log_dir=os.path.expanduser('./logs/'), 
                                   histogram_freq=0, 
                                   #write_batch_performance=True,
                                   write_graph=True,