Feature visualization update (#3920)

detect.py

@@ -40,6 +40,7 @@ def run(weights='yolov5s.pt',  # model.pt path(s)
         classes=None,  # filter by class: --class 0, or --class 0 2 3
         agnostic_nms=False,  # class-agnostic NMS
         augment=False,  # augmented inference
+        visualize=False,  # visualize features
         update=False,  # update all models
         project='runs/detect',  # save results to project/name
         name='exp',  # save results to project/name
@@ -100,7 +101,9 @@ def run(weights='yolov5s.pt',  # model.pt path(s)
 
         # Inference
         t1 = time_synchronized()
-        pred = model(img, augment=augment)[0]
+        pred = model(img,
+                     augment=augment,
+                     visualize=increment_path(save_dir / 'features', mkdir=True) if visualize else False)[0]
 
         # Apply NMS
         pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det)
@@ -201,6 +204,7 @@ def parse_opt():
     parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
     parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
     parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--visualize', action='store_true', help='visualize features')
     parser.add_argument('--update', action='store_true', help='update all models')
     parser.add_argument('--project', default='runs/detect', help='save results to project/name')
     parser.add_argument('--name', default='exp', help='save results to project/name')

models/yolo.py

@@ -117,11 +117,10 @@ class Model(nn.Module):
         self.info()
         logger.info('')
 
-    def forward(self, x, augment=False, profile=False):
+    def forward(self, x, augment=False, profile=False, visualize=False):
         if augment:
             return self.forward_augment(x)  # augmented inference, None
-        else:
-            return self.forward_once(x, profile)  # single-scale inference, train
+        return self.forward_once(x, profile, visualize)  # single-scale inference, train
 
     def forward_augment(self, x):
         img_size = x.shape[-2:]  # height, width
@@ -136,7 +135,7 @@ class Model(nn.Module):
             y.append(yi)
         return torch.cat(y, 1), None  # augmented inference, train
 
-    def forward_once(self, x, profile=False, feature_vis=False):
+    def forward_once(self, x, profile=False, visualize=False):
         y, dt = [], []  # outputs
         for m in self.model:
             if m.f != -1:  # if not from previous layer
@@ -155,8 +154,8 @@ class Model(nn.Module):
             x = m(x)  # run
             y.append(x if m.i in self.save else None)  # save output
 
-            if feature_vis and m.type == 'models.common.SPP':
-                feature_visualization(x, m.type, m.i)
+            if visualize:
+                feature_visualization(x, m.type, m.i, save_dir=visualize)
 
         if profile:
             logger.info('%.1fms total' % sum(dt))

utils/plots.py

 # Plotting utils
 
 import glob
-import math
 import os
 from copy import copy
 from pathlib import Path
 
 import cv2
+import math
 import matplotlib
 import matplotlib.pyplot as plt
 import numpy as np
@@ -15,7 +15,6 @@ import seaborn as sn
 import torch
 import yaml
 from PIL import Image, ImageDraw, ImageFont
-from torchvision import transforms
 
 from utils.general import increment_path, xywh2xyxy, xyxy2xywh
 from utils.metrics import fitness
@@ -448,28 +447,26 @@ def plot_results(start=0, stop=0, bucket='', id=(), labels=(), save_dir=''):
     fig.savefig(Path(save_dir) / 'results.png', dpi=200)
 
 
-def feature_visualization(x, module_type, stage, n=64):
+def feature_visualization(x, module_type, stage, n=64, save_dir=Path('runs/detect/exp')):
     """
     x:              Features to be visualized
     module_type:    Module type
     stage:          Module stage within model
     n:              Maximum number of feature maps to plot
+    save_dir:       Directory to save results
     """
-    batch, channels, height, width = x.shape  # batch, channels, height, width
-    if height > 1 and width > 1:
-        project, name = 'runs/features', 'exp'
-        save_dir = increment_path(Path(project) / name)  # increment run
-        save_dir.mkdir(parents=True, exist_ok=True)  # make dir
-
-        plt.figure(tight_layout=True)
-        blocks = torch.chunk(x, channels, dim=1)  # block by channel dimension
-        n = min(n, len(blocks))
-        for i in range(n):
-            feature = transforms.ToPILImage()(blocks[i].squeeze())
-            ax = plt.subplot(int(math.sqrt(n)), int(math.sqrt(n)), i + 1)
-            ax.axis('off')
-            plt.imshow(feature)  # cmap='gray'
-
-        f = f"stage_{stage}_{module_type.split('.')[-1]}_features.png"
-        print(f'Saving {save_dir / f}...')
-        plt.savefig(save_dir / f, dpi=300)
+    if 'Detect' not in module_type:
+        batch, channels, height, width = x.shape  # batch, channels, height, width
+        if height > 1 and width > 1:
+            f = f"stage{stage}_{module_type.split('.')[-1]}_features.png"  # filename
+
+            plt.figure(tight_layout=True)
+            blocks = torch.chunk(x[0], channels, dim=0)  # select batch index 0, block by channels
+            n = min(n, channels)  # number of plots
+            ax = plt.subplots(math.ceil(n / 8), 8, tight_layout=True)[1].ravel()  # 8 rows x n/8 cols
+            for i in range(n):
+                ax[i].imshow(blocks[i].squeeze())  # cmap='gray'
+                ax[i].axis('off')
+
+            print(f'Saving {save_dir / f}... ({n}/{channels})')
+            plt.savefig(save_dir / f, dpi=300)