move boolean string to utils

Author: zylo117

coco_eval.py

@@ -22,17 +22,17 @@
 
 from backbone import EfficientDetBackbone
 from efficientdet.utils import BBoxTransform, ClipBoxes
-from utils.utils import preprocess, invert_affine, postprocess
+from utils.utils import preprocess, invert_affine, postprocess, boolean_string
 
 ap = argparse.ArgumentParser()
 ap.add_argument('-p', '--project', type=str, default='coco', help='project file that contains parameters')
 ap.add_argument('-c', '--compound_coef', type=int, default=0, help='coefficients of efficientdet')
 ap.add_argument('-w', '--weights', type=str, default=None, help='/path/to/weights')
 ap.add_argument('--nms_threshold', type=float, default=0.5, help='nms threshold, don\'t change it if not for testing purposes')
-ap.add_argument('--cuda', type=bool, default=True)
+ap.add_argument('--cuda', type=boolean_string, default=True)
 ap.add_argument('--device', type=int, default=0)
-ap.add_argument('--float16', type=bool, default=False)
-ap.add_argument('--override', type=bool, default=True, help='override previous bbox results file if exists')
+ap.add_argument('--float16', type=boolean_string, default=False)
+ap.add_argument('--override', type=boolean_string, default=True, help='override previous bbox results file if exists')
 args = ap.parse_args()
 
 compound_coef = args.compound_coef
@@ -49,7 +49,7 @@
 params = yaml.safe_load(open(f'projects/{project_name}.yml'))
 obj_list = params['obj_list']
 
-input_sizes = [512, 640, 768, 896, 1024, 1280, 1280, 1536]
+input_sizes = [512, 640, 768, 896, 1024, 1280, 1280, 1536, 1536]
 
 
 def evaluate_coco(img_path, set_name, image_ids, coco, model, threshold=0.05):

train.py

@@ -2,25 +2,25 @@
 # adapted from https://github.com/signatrix/efficientdet/blob/master/train.py
 # modified by Zylo117
 
+import argparse
 import datetime
 import os
-import argparse
 import traceback
 
+import numpy as np
 import torch
 import yaml
+from tensorboardX import SummaryWriter
 from torch import nn
 from torch.utils.data import DataLoader
 from torchvision import transforms
-from efficientdet.dataset import CocoDataset, Resizer, Normalizer, Augmenter, collater
-from backbone import EfficientDetBackbone
-from tensorboardX import SummaryWriter
-import numpy as np
 from tqdm.autonotebook import tqdm
 
+from backbone import EfficientDetBackbone
+from efficientdet.dataset import CocoDataset, Resizer, Normalizer, Augmenter, collater
 from efficientdet.loss import FocalLoss
 from utils.sync_batchnorm import patch_replication_callback
-from utils.utils import replace_w_sync_bn, CustomDataParallel, get_last_weights, init_weights
+from utils.utils import replace_w_sync_bn, CustomDataParallel, get_last_weights, init_weights, boolean_string
 
 
 class Params:
@@ -56,19 +56,14 @@ def get_args():
     parser.add_argument('-w', '--load_weights', type=str, default=None,
                         help='whether to load weights from a checkpoint, set None to initialize, set \'last\' to load last checkpoint')
     parser.add_argument('--saved_path', type=str, default='logs/')
-    parser.add_argument('--debug', type=boolean_string, default=False, help='whether visualize the predicted boxes of training, '
-                                                                  'the output images will be in test/')
+    parser.add_argument('--debug', type=boolean_string, default=False,
+                        help='whether visualize the predicted boxes of training, '
+                             'the output images will be in test/')
 
     args = parser.parse_args()
     return args
 
 
-def boolean_string(s):
-    if s not in {'False', 'True'}:
-        raise ValueError('Not a valid boolean string')
-    return s == 'True'
-
-
 class ModelWithLoss(nn.Module):
     def __init__(self, model, debug=False):
         super().__init__()
@@ -308,7 +303,7 @@ def freeze_backbone(m):
                     save_checkpoint(model, f'efficientdet-d{opt.compound_coef}_{epoch}_{step}.pth')
 
                 model.train()
-                           
+
                 # Early stopping
                 if epoch - best_epoch > opt.es_patience > 0:
                     print('[Info] Stop training at epoch {}. The lowest loss achieved is {}'.format(epoch, best_loss))

utils/utils.py

@@ -1,22 +1,21 @@
 # Author: Zylo117
 
+import math
 import os
+import uuid
+from glob import glob
+from typing import Union
 
 import cv2
 import numpy as np
 import torch
-from glob import glob
+import webcolors
 from torch import nn
-from torchvision.ops import nms
+from torch.nn.init import _calculate_fan_in_and_fan_out, _no_grad_normal_
 from torchvision.ops.boxes import batched_nms
-from typing import Union
-import uuid
 
 from utils.sync_batchnorm import SynchronizedBatchNorm2d
 
-from torch.nn.init import _calculate_fan_in_and_fan_out, _no_grad_normal_
-import math
-import webcolors
 
 def invert_affine(metas: Union[float, list, tuple], preds):
     for i in range(len(preds)):
@@ -245,7 +244,7 @@ def variance_scaling_(tensor, gain=1.):
 
 
 STANDARD_COLORS = [
-    'LawnGreen', 'Chartreuse', 'Aqua','Beige', 'Azure','BlanchedAlmond','Bisque',
+    'LawnGreen', 'Chartreuse', 'Aqua', 'Beige', 'Azure', 'BlanchedAlmond', 'Bisque',
     'Aquamarine', 'BlueViolet', 'BurlyWood', 'CadetBlue', 'AntiqueWhite',
     'Chocolate', 'Coral', 'CornflowerBlue', 'Cornsilk', 'Crimson', 'Cyan',
     'DarkCyan', 'DarkGoldenRod', 'DarkGrey', 'DarkKhaki', 'DarkOrange',
@@ -272,14 +271,14 @@ def variance_scaling_(tensor, gain=1.):
 
 
 def from_colorname_to_bgr(color):
-    rgb_color=webcolors.name_to_rgb(color)
-    result=(rgb_color.blue,rgb_color.green,rgb_color.red)
+    rgb_color = webcolors.name_to_rgb(color)
+    result = (rgb_color.blue, rgb_color.green, rgb_color.red)
     return result
 
 
 def standard_to_bgr(list_color_name):
-    standard= []
-    for i in range(len(list_color_name)-36): #-36 used to match the len(obj_list)
+    standard = []
+    for i in range(len(list_color_name) - 36):  # -36 used to match the len(obj_list)
         standard.append(from_colorname_to_bgr(list_color_name[i]))
     return standard
 
@@ -296,11 +295,18 @@ def plot_one_box(img, coord, label=None, score=None, color=None, line_thickness=
     cv2.rectangle(img, c1, c2, color, thickness=tl)
     if label:
         tf = max(tl - 2, 1)  # font thickness
-        s_size = cv2.getTextSize(str('{:.0%}'.format(score)),0, fontScale=float(tl) / 3, thickness=tf)[0]
+        s_size = cv2.getTextSize(str('{:.0%}'.format(score)), 0, fontScale=float(tl) / 3, thickness=tf)[0]
         t_size = cv2.getTextSize(label, 0, fontScale=float(tl) / 3, thickness=tf)[0]
-        c2 = c1[0] + t_size[0]+s_size[0]+15, c1[1] - t_size[1] -3
-        cv2.rectangle(img, c1, c2 , color, -1)  # filled
-        cv2.putText(img, '{}: {:.0%}'.format(label, score), (c1[0],c1[1] - 2), 0, float(tl) / 3, [0, 0, 0], thickness=tf, lineType=cv2.FONT_HERSHEY_SIMPLEX)
+        c2 = c1[0] + t_size[0] + s_size[0] + 15, c1[1] - t_size[1] - 3
+        cv2.rectangle(img, c1, c2, color, -1)  # filled
+        cv2.putText(img, '{}: {:.0%}'.format(label, score), (c1[0], c1[1] - 2), 0, float(tl) / 3, [0, 0, 0],
+                    thickness=tf, lineType=cv2.FONT_HERSHEY_SIMPLEX)
+
 
-        
 color_list = standard_to_bgr(STANDARD_COLORS)
+
+
+def boolean_string(s):
+    if s not in {'False', 'True'}:
+        raise ValueError('Not a valid boolean string')
+    return s == 'True'