python 3 support, updated readme, setup.py for building extensions, other minor changes

.gitignore

@@ -0,0 +1,5 @@
+.idea/
+build/
+data/
+__pycache__
+output

README.md

@@ -1,7 +1,7 @@
 # High-level Semantic Feature Detection: A New Perspective for Pedestrian Detection
 Keras implementation of [CSP] accepted by CVPR 2019.
 ## Introduction
-This paper provides a new perspective for detecting pedestrians where detection is formulated as Center and Scale Prediction (CSP), the pipeline is illustrated in the following. For more details, please refer to our [paper](./docs/2019CVPR-CSP.pdf).
+This paper provides a new perspective for detecting pedestrians where detection is formulated as Center and Scale Prediction (CSP), the pipeline is illustrated in the following. For more details, please refer to our [paper](http://openaccess.thecvf.com/content_CVPR_2019/papers/Liu_High-Level_Semantic_Feature_Detection_A_New_Perspective_for_Pedestrian_Detection_CVPR_2019_paper.pdf).
 ![img01](./docs/pipeline.png)
 
 Besides the superority on pedestrian detection demonstrated in the paper, we take a step further towards the generablity of CSP and validate it on face detection. Experimental reults on WiderFace benchmark also show the competitiveness of CSP.
@@ -10,10 +10,10 @@ Besides the superority on pedestrian detection demonstrated in the paper, we tak
 
 ### Dependencies
 
-* Python 2.7
-* Tensorflow 1.4.1
-* Keras 2.0.6
-* OpenCV 3.4.1.15
+* Python >= 3.6
+* Tensorflow >= 1.1.3
+* Keras >= 2.0.6
+* OpenCV >= 3.4.1.15  (note that other versions than 3.4.1.15 will result in different performance on Caltech)
 
 ## Contents
 1. [Installation](#installation)
@@ -33,6 +33,17 @@ Besides the superority on pedestrian detection demonstrated in the paper, we tak
   pip install -r requirements.txt
 ```
 
+3. Build dependencies
+```
+  python setup.py build_ext --inplace
+```
+
+4. Download pretrained resnet50 weights (basenet only):
+```
+./download_weights.sh
+```
+
+
 ### Preparation
 1. Download the dataset.
 

download_weights.sh

@@ -0,0 +1,2 @@
+mkdir -p data/models
+wget -O ./data/models/resnet50_weights_tf_dim_ordering_tf_kernels.h5 https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels.h5 --no-check-certificate

eval_city/cocoapi/PythonAPI/pycocotools/coco.py

@@ -58,7 +58,7 @@
 import sys
 PYTHON_VERSION = sys.version_info[0]
 if PYTHON_VERSION == 2:
-    from urllib import urlretrieve
+    from urllib.request import urlretrieve
 elif PYTHON_VERSION == 3:
     from urllib.request import urlretrieve
 
@@ -83,7 +83,7 @@ def __init__(self, annotation_file=None):
             tic = time.time()
             dataset = json.load(open(annotation_file, 'r'))
             assert type(dataset)==dict, 'annotation file format {} not supported'.format(type(dataset))
-            print('Done (t={:0.2f}s)'.format(time.time()- tic))
+            print(('Done (t={:0.2f}s)'.format(time.time()- tic)))
             self.dataset = dataset
             self.createIndex()
 
@@ -123,8 +123,8 @@ def info(self):
         Print information about the annotation file.
         :return:
         """
-        for key, value in self.dataset['info'].items():
-            print('{}: {}'.format(key, value))
+        for key, value in list(self.dataset['info'].items()):
+            print(('{}: {}'.format(key, value)))
 
     def getAnnIds(self, imgIds=[], catIds=[], areaRng=[], iscrowd=None):
         """
@@ -187,7 +187,7 @@ def getImgIds(self, imgIds=[], catIds=[]):
         catIds = catIds if _isArrayLike(catIds) else [catIds]
 
         if len(imgIds) == len(catIds) == 0:
-            ids = self.imgs.keys()
+            ids = list(self.imgs.keys())
         else:
             ids = set(imgIds)
             for i, catId in enumerate(catIds):
@@ -292,7 +292,7 @@ def showAnns(self, anns):
             ax.add_collection(p)
         elif datasetType == 'captions':
             for ann in anns:
-                print(ann['caption'])
+                print((ann['caption']))
 
     def loadRes(self, resFile):
         """
@@ -305,7 +305,7 @@ def loadRes(self, resFile):
 
         print('Loading and preparing results...')
         tic = time.time()
-        if type(resFile) == str or type(resFile) == unicode:
+        if type(resFile) == str or type(resFile) == str:
             anns = json.load(open(resFile))
         elif type(resFile) == np.ndarray:
             anns = self.loadNumpyAnnotations(resFile)
@@ -349,7 +349,7 @@ def loadRes(self, resFile):
                 ann['area'] = (x1-x0)*(y1-y0)
                 ann['id'] = id + 1
                 ann['bbox'] = [x0,y0,x1-x0,y1-y0]
-        print('DONE (t={:0.2f}s)'.format(time.time()- tic))
+        print(('DONE (t={:0.2f}s)'.format(time.time()- tic)))
 
         res.dataset['annotations'] = anns
         res.createIndex()
@@ -366,7 +366,7 @@ def download(self, tarDir = None, imgIds = [] ):
             print('Please specify target directory')
             return -1
         if len(imgIds) == 0:
-            imgs = self.imgs.values()
+            imgs = list(self.imgs.values())
         else:
             imgs = self.loadImgs(imgIds)
         N = len(imgs)
@@ -377,7 +377,7 @@ def download(self, tarDir = None, imgIds = [] ):
             fname = os.path.join(tarDir, img['file_name'])
             if not os.path.exists(fname):
                 urlretrieve(img['coco_url'], fname)
-            print('downloaded {}/{} images (t={:0.1f}s)'.format(i, N, time.time()- tic))
+            print(('downloaded {}/{} images (t={:0.1f}s)'.format(i, N, time.time()- tic)))
 
     def loadNumpyAnnotations(self, data):
         """
@@ -387,13 +387,13 @@ def loadNumpyAnnotations(self, data):
         """
         print('Converting ndarray to lists...')
         assert(type(data) == np.ndarray)
-        print(data.shape)
+        print((data.shape))
         assert(data.shape[1] == 7)
         N = data.shape[0]
         ann = []
         for i in range(N):
             if i % 1000000 == 0:
-                print('{}/{}'.format(i,N))
+                print(('{}/{}'.format(i,N)))
             ann += [{
                 'image_id'  : int(data[i, 0]),
                 'bbox'  : [ data[i, 1], data[i, 2], data[i, 3], data[i, 4] ],

eval_city/cocoapi/PythonAPI/pycocotools/cocoeval.py

@@ -130,8 +130,8 @@ def evaluate(self):
         # add backward compatibility if useSegm is specified in params
         if not p.useSegm is None:
             p.iouType = 'segm' if p.useSegm == 1 else 'bbox'
-            print('useSegm (deprecated) is not None. Running {} evaluation'.format(p.iouType))
-        print('Evaluate annotation type *{}*'.format(p.iouType))
+            print(('useSegm (deprecated) is not None. Running {} evaluation'.format(p.iouType)))
+        print(('Evaluate annotation type *{}*'.format(p.iouType)))
         p.imgIds = list(np.unique(p.imgIds))
         if p.useCats:
             p.catIds = list(np.unique(p.catIds))
@@ -159,7 +159,7 @@ def evaluate(self):
              ]
         self._paramsEval = copy.deepcopy(self.params)
         toc = time.time()
-        print('DONE (t={:0.2f}s).'.format(toc-tic))
+        print(('DONE (t={:0.2f}s).'.format(toc-tic)))
 
     def computeIoU(self, imgId, catId):
         p = self.params
@@ -346,7 +346,7 @@ def accumulate(self, p = None):
         # get inds to evaluate
         k_list = [n for n, k in enumerate(p.catIds)  if k in setK]
         m_list = [m for n, m in enumerate(p.maxDets) if m in setM]
-        a_list = [n for n, a in enumerate(map(lambda x: tuple(x), p.areaRng)) if a in setA]
+        a_list = [n for n, a in enumerate([tuple(x) for x in p.areaRng]) if a in setA]
         i_list = [n for n, i in enumerate(p.imgIds)  if i in setI]
         I0 = len(_pe.imgIds)
         A0 = len(_pe.areaRng)
@@ -418,7 +418,7 @@ def accumulate(self, p = None):
             'scores': scores,
         }
         toc = time.time()
-        print('DONE (t={:0.2f}s).'.format( toc-tic))
+        print(('DONE (t={:0.2f}s).'.format( toc-tic)))
 
     def summarize(self):
         '''
@@ -454,7 +454,7 @@ def _summarize( ap=1, iouThr=None, areaRng='all', maxDets=100 ):
                 mean_s = -1
             else:
                 mean_s = np.mean(s[s>-1])
-            print(iStr.format(titleStr, typeStr, iouStr, areaRng, maxDets, mean_s))
+            print((iStr.format(titleStr, typeStr, iouStr, areaRng, maxDets, mean_s)))
             return mean_s
         def _summarizeDets():
             stats = np.zeros((12,))

eval_city/eval_script/coco.py

@@ -58,7 +58,7 @@
 import sys
 PYTHON_VERSION = sys.version_info[0]
 if PYTHON_VERSION == 2:
-    from urllib import urlretrieve
+    from urllib.request import urlretrieve
 elif PYTHON_VERSION == 3:
     from urllib.request import urlretrieve
 
@@ -118,8 +118,8 @@ def info(self):
         Print information about the annotation file.
         :return:
         """
-        for key, value in self.dataset['info'].items():
-            print('{}: {}'.format(key, value))
+        for key, value in list(self.dataset['info'].items()):
+            print(('{}: {}'.format(key, value)))
 
     def getAnnIds(self, imgIds=[], catIds=[], areaRng=[], iscrowd=None):
         """
@@ -182,7 +182,7 @@ def getImgIds(self, imgIds=[], catIds=[]):
         catIds = catIds if type(catIds) == list else [catIds]
 
         if len(imgIds) == len(catIds) == 0:
-            ids = self.imgs.keys()
+            ids = list(self.imgs.keys())
         else:
             ids = set(imgIds)
             for i, catId in enumerate(catIds):
@@ -287,7 +287,7 @@ def showAnns(self, anns):
             ax.add_collection(p)
         elif datasetType == 'captions':
             for ann in anns:
-                print(ann['caption'])
+                print((ann['caption']))
 
     def loadRes(self, resFile):
         """
@@ -300,7 +300,7 @@ def loadRes(self, resFile):
 
         # print('Loading and preparing results...')
         tic = time.time()
-        if type(resFile) == str or type(resFile) == unicode:
+        if type(resFile) == str or type(resFile) == str:
             anns = json.load(open(resFile))
         elif type(resFile) == np.ndarray:
             anns = self.loadNumpyAnnotations(resFile)
@@ -364,7 +364,7 @@ def download(self, tarDir = None, imgIds = [] ):
             print('Please specify target directory')
             return -1
         if len(imgIds) == 0:
-            imgs = self.imgs.values()
+            imgs = list(self.imgs.values())
         else:
             imgs = self.loadImgs(imgIds)
         N = len(imgs)
@@ -375,7 +375,7 @@ def download(self, tarDir = None, imgIds = [] ):
             fname = os.path.join(tarDir, img['file_name'])
             if not os.path.exists(fname):
                 urlretrieve(img['coco_url'], fname)
-            print('downloaded {}/{} images (t={:0.1f}s)'.format(i, N, time.time()- tic))
+            print(('downloaded {}/{} images (t={:0.1f}s)'.format(i, N, time.time()- tic)))
 
     def loadNumpyAnnotations(self, data):
         """
@@ -385,13 +385,13 @@ def loadNumpyAnnotations(self, data):
         """
         print('Converting ndarray to lists...')
         assert(type(data) == np.ndarray)
-        print(data.shape)
+        print((data.shape))
         assert(data.shape[1] == 7)
         N = data.shape[0]
         ann = []
         for i in range(N):
             if i % 1000000 == 0:
-                print('{}/{}'.format(i,N))
+                print(('{}/{}'.format(i,N)))
             ann += [{
                 'image_id'  : int(data[i, 0]),
                 'bbox'  : [ data[i, 1], data[i, 2], data[i, 3], data[i, 4] ],

eval_city/eval_script/eval_MR_multisetup.py

@@ -122,7 +122,7 @@ def evaluate(self, id_setup):
         # add backward compatibility if useSegm is specified in params
         if not p.useSegm is None:
             p.iouType = 'segm' if p.useSegm == 1 else 'bbox'
-            print('useSegm (deprecated) is not None. Running {} evaluation'.format(p.iouType))
+            print(('useSegm (deprecated) is not None. Running {} evaluation'.format(p.iouType)))
         # print('Evaluate annotation type *{}*'.format(p.iouType))
         p.imgIds = list(np.unique(p.imgIds))
         if p.useCats:
@@ -446,7 +446,7 @@ def _summarize(iouThr=None, maxDets=100 ):
                 mean_s = np.log(mrs[mrs<2])
                 mean_s = np.mean(mean_s)
                 mean_s = np.exp(mean_s)
-            print(iStr.format(titleStr, typeStr,setupStr, iouStr, heightStr, occlStr, mean_s*100))
+            print((iStr.format(titleStr, typeStr,setupStr, iouStr, heightStr, occlStr, mean_s*100)))
             # res_file.write(iStr.format(titleStr, typeStr,setupStr, iouStr, heightStr, occlStr, mean_s*100))
             res_file.write(str(mean_s * 100))
             res_file.write('\n')

eval_city/eval_script/eval_demo.py

@@ -8,7 +8,7 @@
 annFile = '../val_gt.json'
 main_path = '../../output/valresults/city/h/off'
 for f in sorted(os.listdir(main_path)):
-    print f
+    print(f)
     # initialize COCO detections api
     dt_path = os.path.join(main_path, f)
     resFile = os.path.join(dt_path,'val_dt.json')

generate_cache_caltech.py

@@ -1,5 +1,5 @@
 import os
-import cPickle
+import pickle
 import numpy as np
 import matplotlib.pyplot as plt
 
@@ -18,46 +18,47 @@
 box_count = 0
 files = sorted(os.listdir(all_anno_path))
 for l in range(len(files)):
-	gtname = files[l]
-	imgname = files[l].split('.')[0]+'.jpg'
-	img_path = os.path.join(all_img_path, imgname)
-	gt_path = os.path.join(all_anno_path, gtname)
+    gtname = files[l]
+    imgname = files[l].split('.')[0] + '.jpg'
+    img_path = os.path.join(all_img_path, imgname)
+    gt_path = os.path.join(all_anno_path, gtname)
 
-	boxes = []
-	ig_boxes = []
-	with open(gt_path, 'rb') as fid:
-		lines = fid.readlines()
-	if len(lines)>1:
-		for i in range(1, len(lines)):
-			info = lines[i].strip().split(' ')
-			label = info[0]
-			occ, ignore = info[5], info[10]
-			x1, y1 = max(int(float(info[1])), 0), max(int(float(info[2])), 0)
-			w, h = min(int(float(info[3])), cols - x1 - 1), min(int(float(info[4])), rows - y1 - 1)
-			box = np.array([int(x1), int(y1), int(x1) + int(w), int(y1) + int(h)])
-			if int(ignore) == 0:
-				boxes.append(box)
-			else:
-				ig_boxes.append(box)
-	boxes = np.array(boxes)
-	ig_boxes = np.array(ig_boxes)
+    boxes = []
+    ig_boxes = []
+    with open(gt_path, 'rb') as fid:
+        lines = fid.readlines()
+    if len(lines) > 1:
+        for i in range(1, len(lines)):
+            info = lines[i].strip().split(' ')
+            label = info[0]
+            occ, ignore = info[5], info[10]
+            x1, y1 = max(int(float(info[1])), 0), max(int(float(info[2])), 0)
+            w, h = min(int(float(info[3])), cols - x1 - 1), min(int(float(info[4])), rows - y1 - 1)
+            box = np.array([int(x1), int(y1), int(x1) + int(w), int(y1) + int(h)])
+            if int(ignore) == 0:
+                boxes.append(box)
+            else:
+                ig_boxes.append(box)
+    boxes = np.array(boxes)
+    ig_boxes = np.array(ig_boxes)
 
-	annotation = {}
-	annotation['filepath'] = img_path
-	box_count += len(boxes)
-	iggt_count += len(ig_boxes)
-	annotation['bboxes'] = boxes
-	annotation['ignoreareas'] = ig_boxes
-	if len(boxes) == 0:
-		image_data_nogt.append(annotation)
-	else:
-		image_data_gt.append(annotation)
-		valid_count += 1
-print '{} images and {} valid images, {} valid gt and {} ignored gt'.format(len(files), valid_count, box_count, iggt_count)
+    annotation = {}
+    annotation['filepath'] = img_path
+    box_count += len(boxes)
+    iggt_count += len(ig_boxes)
+    annotation['bboxes'] = boxes
+    annotation['ignoreareas'] = ig_boxes
+    if len(boxes) == 0:
+        image_data_nogt.append(annotation)
+    else:
+        image_data_gt.append(annotation)
+        valid_count += 1
+print('{} images and {} valid images, {} valid gt and {} ignored gt'.format(len(files), valid_count, box_count,
+                                                                            iggt_count))
 
 if not os.path.exists(res_path_gt):
-	with open(res_path_gt, 'wb') as fid:
-		cPickle.dump(image_data_gt, fid, cPickle.HIGHEST_PROTOCOL)
+    with open(res_path_gt, 'wb') as fid:
+        pickle.dump(image_data_gt, fid, pickle.HIGHEST_PROTOCOL)
 if not os.path.exists(res_path_nogt):
-	with open(res_path_nogt, 'wb') as fid:
-		cPickle.dump(image_data_nogt, fid, cPickle.HIGHEST_PROTOCOL)
+    with open(res_path_nogt, 'wb') as fid:
+        pickle.dump(image_data_nogt, fid, pickle.HIGHEST_PROTOCOL)