refine infer with args (#2501)

* refine infer with args

* remove samples

* fix as review

* refine code

* refine args

* move get_test_images to infer.py

* add visualize log

* fix images = []

* fix args

* refine infer.py

* fix yolov3_r34.yml

PaddleCV/object_detection/configs/cascade_rcnn_r50_fpn_1x.yml

@@ -136,7 +136,6 @@ FasterRCNNTestFeed:
   dataset:
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   drop_last: false
   num_workers: 2
   shuffle: false

PaddleCV/object_detection/configs/faster_rcnn_r101_1x.yml

@@ -115,5 +115,3 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/faster_rcnn_r101_fpn_1x.yml

@@ -138,6 +138,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_r101_fpn_2x.yml

@@ -138,6 +138,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_r101_vd_fpn_1x.yml

@@ -139,6 +139,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_r101_vd_fpn_2x.yml

@@ -139,6 +139,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_r50_1x.yml

@@ -115,5 +115,3 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/faster_rcnn_r50_2x.yml

@@ -115,5 +115,3 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/faster_rcnn_r50_fpn_1x.yml

@@ -136,7 +136,6 @@ FasterRCNNTestFeed:
   dataset:
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   drop_last: false
   num_workers: 2
   shuffle: false

PaddleCV/object_detection/configs/faster_rcnn_r50_vd_1x.yml

@@ -117,5 +117,3 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/faster_rcnn_r50_vd_fpn_2x.yml

@@ -139,6 +139,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_se154_1x.yml

@@ -124,6 +124,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_se154_fpn_1x.yml

@@ -141,6 +141,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_se154_fpn_s1x.yml

@@ -141,6 +141,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_x101_64x4d_fpn_1x.yml

@@ -140,6 +140,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/faster_rcnn_x101_64x4d_fpn_2x.yml

@@ -140,6 +140,5 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   shuffle: False

PaddleCV/object_detection/configs/mask_rcnn_r101_fpn_1x.yml

@@ -147,7 +147,5 @@ MaskRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5
   num_workers: 2
   use_padded_im_info: True

PaddleCV/object_detection/configs/mask_rcnn_r101_fpn_2x.yml

@@ -147,7 +147,5 @@ MaskRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5
   num_workers: 2
   use_padded_im_info: True

PaddleCV/object_detection/configs/mask_rcnn_r50_1x.yml

@@ -128,5 +128,3 @@ MaskRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/mask_rcnn_r50_2x.yml

@@ -129,5 +129,3 @@ MaskRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/mask_rcnn_r50_fpn_1x.yml

@@ -147,7 +147,5 @@ MaskRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5
   num_workers: 2
   use_padded_im_info: True

PaddleCV/object_detection/configs/mask_rcnn_r50_fpn_2x.yml

@@ -147,7 +147,5 @@ MaskRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
-  samples: 5
   num_workers: 2
   use_padded_im_info: True

PaddleCV/object_detection/configs/mask_rcnn_r50_vd_fpn_2x.yml

@@ -150,6 +150,5 @@ MaskRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   num_workers: 2
   use_padded_im_info: True

PaddleCV/object_detection/configs/retinanet_r50_fpn_1x.yml

@@ -151,7 +151,6 @@ FasterRCNNTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: val2017.txt
   drop_last: false
   image_shape: [3, 1333, 800]
   num_workers: 2

PaddleCV/object_detection/configs/ssd_mobilenet_v1_voc.yml

@@ -84,4 +84,3 @@ SSDTestFeed:
     image_dir: VOCdevkit/VOC_all/JPEGImages
     use_default_label: false
   drop_last: false
-  test_file: data/voc/VOCdevkit/VOC_all/ImageSets/Main/test.txt

PaddleCV/object_detection/configs/yolov3_darknet.yml

@@ -80,5 +80,3 @@ YoloTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: ../val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/yolov3_mobilenet_v1.yml

@@ -81,5 +81,3 @@ YoloTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: ../val2017.txt
-  samples: 5

PaddleCV/object_detection/configs/yolov3_r34.yml

@@ -17,6 +17,8 @@ YOLOv3:
 
 ResNet:
   norm_type: sync_bn
+  freeze_at: 0
+  freeze_norm: False
   norm_decay: 0.
   depth: 34
   feature_maps: [3, 4, 5]
@@ -81,5 +83,3 @@ YoloTestFeed:
     dataset_dir: data/coco
     annotation: annotations/instances_val2017.json
     image_dir: val2017
-  test_file: ../val2017.txt
-  samples: 5

PaddleCV/object_detection/ppdet/data/data_feed.py

@@ -82,8 +82,8 @@ def create_reader(feed, max_iter=0):
         }
     }
     
-    if mode == 'TEST':
-        data_config[mode]['TEST_FILE'] = feed.test_file
+    if len(getattr(feed.dataset, 'images', [])) > 0:
+        data_config[mode]['IMAGES'] = feed.dataset.images
 
     transform_config = {
         'WORKER_CONF': {
@@ -244,12 +244,16 @@ class SimpleDataSet(DataSet):
     __source__ = 'SimpleSource'
 
     def __init__(self,
-                 dataset_dir=VOC_DATASET_DIR,
-                 annotation=VOC_TEST_ANNOTATION,
-                 image_dir=VOC_IMAGE_DIR,
-                 use_default_label=VOC_USE_DEFAULT_LABEL):
+                 dataset_dir=None,
+                 annotation=None,
+                 image_dir=None,
+                 use_default_label=None):
         super(SimpleDataSet, self).__init__(
             dataset_dir=dataset_dir, annotation=annotation, image_dir=image_dir)
+        self.images = []
+
+    def add_images(self, images):
+        self.images.extend(images)
 
 
 @serializable
@@ -281,7 +285,6 @@ class DataFeed(object):
                  samples=-1,
                  drop_last=False,
                  with_background=True,
-                 test_file=None,
                  num_workers=2,
                  bufsize=10,
                  use_process=False,
@@ -296,7 +299,6 @@ class DataFeed(object):
         self.samples = samples
         self.drop_last = drop_last
         self.with_background = with_background
-        self.test_file = test_file
         self.num_workers = num_workers
         self.bufsize = bufsize
         self.use_process = use_process
@@ -385,7 +387,6 @@ class TestFeed(DataFeed):
                  shuffle=False,
                  drop_last=False,
                  with_background=True,
-                 test_file=None,
                  num_workers=2):
         super(TestFeed, self).__init__(
             dataset,
@@ -397,7 +398,6 @@ class TestFeed(DataFeed):
             shuffle=shuffle,
             drop_last=drop_last,
             with_background=with_background,
-            test_file=test_file,
             num_workers=num_workers)
 
 
@@ -522,7 +522,6 @@ class FasterRCNNEvalFeed(DataFeed):
                  shuffle=False,
                  samples=-1,
                  drop_last=False,
-                 test_file=None,
                  num_workers=2,
                  use_padded_im_info=True):
         sample_transforms.append(ArrangeTestRCNN())
@@ -536,7 +535,6 @@ class FasterRCNNEvalFeed(DataFeed):
             shuffle=shuffle,
             samples=samples,
             drop_last=drop_last,
-            test_file=test_file,
             num_workers=num_workers,
             use_padded_im_info=use_padded_im_info)
         self.mode = 'VAL'
@@ -564,7 +562,6 @@ class FasterRCNNTestFeed(DataFeed):
                  shuffle=False,
                  samples=-1,
                  drop_last=False,
-                 test_file=None,
                  num_workers=2,
                  use_padded_im_info=True):
         sample_transforms.append(ArrangeTestRCNN())
@@ -580,7 +577,6 @@ class FasterRCNNTestFeed(DataFeed):
             shuffle=shuffle,
             samples=samples,
             drop_last=drop_last,
-            test_file=test_file,
             num_workers=num_workers,
             use_padded_im_info=use_padded_im_info)
         self.mode = 'TEST'
@@ -612,7 +608,6 @@ class MaskRCNNEvalFeed(DataFeed):
                  shuffle=False,
                  samples=-1,
                  drop_last=False,
-                 test_file=None,
                  num_workers=2,
                  use_process=False,
                  use_padded_im_info=True):
@@ -627,7 +622,6 @@ class MaskRCNNEvalFeed(DataFeed):
             shuffle=shuffle,
             samples=samples,
             drop_last=drop_last,
-            test_file=test_file,
             num_workers=num_workers,
             use_process=use_process,
             use_padded_im_info=use_padded_im_info)
@@ -657,7 +651,6 @@ class MaskRCNNTestFeed(DataFeed):
                  shuffle=False,
                  samples=-1,
                  drop_last=False,
-                 test_file=None,
                  num_workers=2,
                  use_process=False,
                  use_padded_im_info=True):
@@ -674,7 +667,6 @@ class MaskRCNNTestFeed(DataFeed):
             shuffle=shuffle,
             samples=samples,
             drop_last=drop_last,
-            test_file=test_file,
             num_workers=num_workers,
             use_process=use_process,
             use_padded_im_info=use_padded_im_info)
@@ -805,7 +797,6 @@ class SSDTestFeed(DataFeed):
                  shuffle=False,
                  samples=-1,
                  drop_last=False,
-                 test_file=None,
                  num_workers=8,
                  bufsize=10,
                  use_process=False):
@@ -822,7 +813,6 @@ class SSDTestFeed(DataFeed):
             shuffle=shuffle,
             samples=samples,
             drop_last=drop_last,
-            test_file=test_file,
             num_workers=num_workers)
         self.mode = 'TEST'
 
@@ -961,10 +951,9 @@ class YoloTestFeed(DataFeed):
                  batch_transforms=[],
                  batch_size=1,
                  shuffle=False,
-                 samples=1,
+                 samples=-1,
                  drop_last=False,
                  with_background=False,
-                 test_file=None,
                  num_workers=8,
                  num_max_boxes=50,
                  use_process=False):
@@ -982,7 +971,6 @@ class YoloTestFeed(DataFeed):
             samples=samples,
             drop_last=drop_last,
             with_background=with_background,
-            test_file=test_file,
             num_workers=num_workers,
             use_process=use_process)
         self.num_max_boxes = num_max_boxes

PaddleCV/object_detection/ppdet/data/source/simple_source.py

@@ -30,26 +30,23 @@ class SimpleSource(Dataset):
     Load image files for testing purpose
 
     Args:
-        test_file (str): list of image file names, relative to `image_dir`
-        image_dir (str): root dir for images
+        images (list): list of path of images
         samples (int): number of samples to load, -1 means all
         load_img (bool): should images be loaded
     """
 
     def __init__(self,
-                 test_file='',
-                 image_dir=None,
+                 images=[],
                  samples=-1,
                  load_img=True,
                  **kwargs):
         super(SimpleSource, self).__init__()
         self._epoch = -1
-        assert test_file != '' and os.path.isfile(test_file), \
-            "test file not found: " + test_file
-        self._fname = test_file
-        self._image_dir = image_dir
-        assert image_dir is not None and os.path.isdir(image_dir), \
-            "image directory not found: " + image_dir
+        for image in images:
+            assert image != '' and os.path.isfile(image), \
+                    "Image {} not found".format(image)
+        self._images = images
+        self._fname = None
         self._simple = None
         self._pos = -1
         self._drained = False
@@ -68,32 +65,25 @@ class SimpleSource(Dataset):
             sample = copy.deepcopy(self._simple[self._pos])
             if self._load_img:
                 sample['image'] = self._load_image(sample['im_file'])
-            else:
-                sample['im_file'] = os.path.join(self._image_dir,
-                                                 sample['im_file'])
+
             self._pos += 1
             return sample
 
     def _load(self):
-        assert os.path.isfile(self._fname) and self._fname.endswith('.txt'), \
-            "invalid test file path"
         ct = 0
         records = []
-        with open(self._fname, 'r') as fr:
-            while True:
-                line = fr.readline().strip()
-                if not line or (self._samples > 0 and ct >= self._samples):
+        for image in self._images:
+            if self._samples > 0 and ct >= self._samples:
                 break
-                rec = {'im_id': np.array([ct]), 'im_file': line}
-                self._imid2path[ct] = line
+            rec = {'im_id': np.array([ct]), 'im_file': image}
+            self._imid2path[ct] = image
             ct += 1
             records.append(rec)
-        assert len(records) > 0, "no image file found in " + self._fname
+        assert len(records) > 0, "no image file found"
         return records
 
     def _load_image(self, where):
-        fn = os.path.join(self._image_dir, where)
-        with open(fn, 'rb') as f:
+        with open(where, 'rb') as f:
             return f.read()
 
     def reset(self):

PaddleCV/object_detection/ppdet/utils/cli.py

@@ -61,6 +61,16 @@ def parse_args():
         action='store_true',
         default=False,
         help="Whether perform evaluation in train")
+    parser.add_argument(
+        "--infer_dir",
+        type=str,
+        default=None,
+        help="Image directory path to perform inference.")
+    parser.add_argument(
+        "--infer_img",
+        type=str,
+        default=None,
+        help="Image path to perform inference, --infer-img has a higher priority than --image-dir")
     parser.add_argument(
         "-o", "--opt", nargs=REMAINDER, help="set configuration options")
     args = parser.parse_args()

PaddleCV/object_detection/ppdet/utils/visualizer.py

@@ -38,22 +38,27 @@ def visualize_results(image_path,
                       bbox_results=None,
                       mask_results=None):
     """
-    TODO(dengkaipeng): add more comments
     Visualize bbox and mask results
     """
-    image = None
+    if not os.path.exists(SAVE_HOME):
+        os.makedirs(SAVE_HOME)
+
+    logger.info("Image {} detect: ".format(image_path))
+    image = Image.open(image_path)
     if mask_results:
-        image = draw_mask(image_path, mask_results, threshold)
+        image = draw_mask(image, mask_results, threshold)
     if bbox_results:
-        draw_bbox(image_path, catid2name, bbox_results, threshold, image)
+        image = draw_bbox(image, catid2name, bbox_results, threshold)
+
+    save_name = get_save_image_name(image_path)
+    logger.info("Detection results save in {}\n".format(save_name))
+    image.save(save_name)
 
 
-def draw_mask(image_path, segms, threshold, alpha=0.7, save_image=False):
+def draw_mask(image, segms, threshold, alpha=0.7):
     """
-    TODO(dengkaipeng): add more comments
     Draw mask on image
     """
-    image = Image.open(image_path)
     im_width, im_height = image.size
     mask_color_id = 0
     w_ratio = .4
@@ -72,23 +77,13 @@ def draw_mask(image_path, segms, threshold, alpha=0.7, save_image=False):
         image[idx[0], idx[1], :] *= 1.0 - alpha
         image[idx[0], idx[1], :] += alpha * color_mask
     image = Image.fromarray(image.astype('uint8'))
-
-    if not os.path.exists(SAVE_HOME):
-        os.makedirs(SAVE_HOME)
-    if save_image:
-        save_name = get_save_image_name(image_path)
-        logger.info("Detection mask results save in {}".format(save_name))
-        image.save(save_name)
     return image
 
 
-def draw_bbox(image_path, catid2name, bboxes, threshold, image=None):
+def draw_bbox(image, catid2name, bboxes, threshold):
     """
-    TODO(dengkaipeng): add more comments
     Draw bbox on image
     """
-    if image is None:
-        image = Image.open(image_path)
     draw = ImageDraw.Draw(image)
     im_width, im_height = image.size
 
@@ -106,13 +101,12 @@ def draw_bbox(image_path, catid2name, bboxes, threshold, image=None):
             fill='red')
         if image.mode == 'RGB':
             draw.text((xmin, ymin), catid2name[catid], (255, 255, 0))
+        logger.info("\t {:15s} at {:25} score: {:.5f}".format(
+                    catid2name[catid], 
+                    str(list(map(int, [xmin, ymin, xmax, ymax]))),
+                    score))
 
-    if not os.path.exists(SAVE_HOME):
-        os.makedirs(SAVE_HOME)
-    save_name = get_save_image_name(image_path)
-    logger.info("Detection bbox results save in {}".format(save_name))
-    image.save(save_name)
-
+    return image
 
 def get_save_image_name(image_path):
     """

PaddleCV/object_detection/tools/infer.py

@@ -17,6 +17,7 @@ from __future__ import division
 from __future__ import print_function
 
 import os
+import glob
 
 import numpy as np
 
@@ -37,6 +38,32 @@ logging.basicConfig(level=logging.INFO, format=FORMAT)
 logger = logging.getLogger(__name__)
 
 
+def get_test_images(infer_dir, infer_img):
+    """
+    Get image path list in TEST mode
+    """
+    assert infer_img is not None or infer_dir is not None, \
+            "--infer-img or --infer-dir should be set"
+    images = []
+
+    # infer_img has a higher priority
+    if infer_img and os.path.isfile(infer_img):
+        images.append(infer_img)
+        return images
+
+    infer_dir = os.path.abspath(infer_dir)
+    assert os.path.isdir(infer_dir), \
+            "infer_dir {} is not a directory".format(infer_dir)
+    for fmt in ['jpg', 'jpeg', 'png', 'bmp']:
+        images.extend(glob.glob('{}/*.{}'.format(infer_dir, fmt)))
+
+    assert len(images) > 0, "no image found in {} with " \
+                "extension {}".format(infer_dir, image_ext)
+    logger.info("Found {} inference images in total.".format(len(images)))
+
+    return images
+
+
 def main():
     args = parse_args()
     cfg = load_config(args.config)
@@ -53,6 +80,9 @@ def main():
     else:
         test_feed = create(cfg['test_feed'])
 
+    test_images = get_test_images(args.infer_dir, args.infer_img)
+    test_feed.dataset.add_images(test_images)
+
     place = fluid.CUDAPlace(0) if cfg['use_gpu'] else fluid.CPUPlace()
     exe = fluid.Executor(place)
 
@@ -103,7 +133,7 @@ def main():
         logger.info('Infer iter {}'.format(iter_id))
 
         im_id = int(res['im_id'][0])
-        image_path = os.path.join(test_feed.dataset.image_dir, imid2path[im_id])
+        image_path = imid2path[im_id]
         if cfg['metric'] == 'COCO':
             bbox_results = None
             mask_results = None