Add compatibility with pifpaf 0.9 (#9)

* remove dependancy on pifpaf.transform

* add compatibility with pifpaf 0.9

* change verbose output

* add image transform compatible with webcam

README.md

@@ -136,9 +136,9 @@ Multiple visualizations can be combined in different windows.
 
 The above gif has been obtained running on a Macbook the command:
 
-`pip3 install opencv-python`
-
-`python3 -m monoloco.run predict --webcam --scale 0.2 --output_types combined --z_max 10 --checkpoint resnet50`
+```pip3 install opencv-python
+python3 -m monoloco.run predict --webcam --scale 0.2 --output_types combined --z_max 10 --checkpoint resnet50 --model data/models/monoloco-190513-1437.pkl
+```
 
 # Preprocess
 

monoloco/__init__.py

@@ -1,4 +1,4 @@
 
 """Open implementation of MonoLoco."""
 
-__version__ = '0.4.3'
+__version__ = '0.4.4'

monoloco/eval/eval_kitti.py

@@ -347,7 +347,7 @@ class EvalKitti:
     def show_statistics(self):
 
         print('-'*90)
-        alp = [[str(100 * average(self.errors[key][perc]))[:4]
+        alp = [[str(100 * average(self.errors[key][perc]))[:5]
                 for perc in ['<0.5m', '<1m', '<2m']]
                for key in self.METHODS]
 
@@ -373,8 +373,8 @@ class EvalKitti:
                     if key == 'our':
                         print("% of annotation inside the confidence interval: {:.1f} %, "
                               "of which {:.1f} % at higher risk"
-                              .format(self.dic_stats['test'][key][clst]['interval'],
-                                      self.dic_stats['test'][key][clst]['at_risk']))
+                              .format(self.dic_stats['test'][key][clst]['interval']*100,
+                                      self.dic_stats['test'][key][clst]['at_risk']*100))
 
                 for perc in ['<0.5m', '<1m', '<2m']:
                     print("{} Instances with error {}: {:.2f} %"

monoloco/network/pifpaf.py

@@ -5,22 +5,18 @@ import numpy as np
 import torchvision
 import torch
 from PIL import Image, ImageFile
-
 from openpifpaf.network import nets
 from openpifpaf import decoder
-from openpifpaf import transforms
+
+from .process import image_transform
 
 
 class ImageList(torch.utils.data.Dataset):
     """It defines transformations to apply to images and outputs of the dataloader"""
-    def __init__(self, image_paths, scale, image_transform=None):
+    def __init__(self, image_paths, scale):
         self.image_paths = image_paths
-        self.image_transform = image_transform or transforms.image_transform  # to_tensor + normalize  (from pifpaf)
         self.scale = scale
 
-        # data = datasets.ImageList(args.images, preprocess=transforms.RescaleRelative(2
-        # .0)
-
     def __getitem__(self, index):
         image_path = self.image_paths[index]
         ImageFile.LOAD_TRUNCATED_IMAGES = True
@@ -34,7 +30,7 @@ class ImageList(torch.utils.data.Dataset):
                                                              interpolation=Image.BICUBIC)
         # PIL images are not iterables
         original_image = torchvision.transforms.functional.to_tensor(image)  # 0-255 --> 0-1
-        image = self.image_transform(image)
+        image = image_transform(image)
 
         return image_path, original_image, image
 

monoloco/network/process.py

@@ -3,6 +3,7 @@ import json
 
 import numpy as np
 import torch
+import torchvision
 
 from ..utils import get_keypoints, pixel_to_camera
 
@@ -82,18 +83,6 @@ def laplace_sampling(outputs, n_samples):
     return xx
 
 
-def epistemic_variance(total_outputs):
-    """Compute epistemic variance"""
-
-    # var_y = np.sum(total_outputs**2, axis=0) / total_outputs.shape[0] - (np.mean(total_outputs, axis=0))**2
-    var_y = np.var(total_outputs, axis=0)
-    lower_b = np.quantile(a=total_outputs, q=0.25, axis=0)
-    upper_b = np.quantile(a=total_outputs, q=0.75, axis=0)
-    var_new = (upper_b - lower_b)
-
-    return var_y, var_new
-
-
 def unnormalize_bi(outputs):
     """Unnormalize relative bi of a nunmpy array"""
 
@@ -151,3 +140,13 @@ def prepare_pif_kps(kps_in):
     ccs = kps_in[2:][::3]
 
     return [xxs, yys, ccs]
+
+
+def image_transform(image):
+
+    normalize = torchvision.transforms.Normalize(
+        mean=[0.485, 0.456, 0.406],
+        std=[0.229, 0.224, 0.225]
+    )
+    transforms = torchvision.transforms.Compose([torchvision.transforms.ToTensor(), normalize, ])
+    return transforms(image)

monoloco/visuals/webcam.py

@@ -11,12 +11,11 @@ import time
 import torch
 import matplotlib.pyplot as plt
 from PIL import Image
-from openpifpaf import transforms
 import cv2
 
 from ..visuals import Printer
 from ..network import PifPaf, MonoLoco
-from ..network.process import preprocess_pifpaf, factory_for_gt
+from ..network.process import preprocess_pifpaf, factory_for_gt, image_transform
 
 
 def webcam(args):
@@ -42,7 +41,7 @@ def webcam(args):
         height, width, _ = image.shape
         print('resized image size: {}'.format(image.shape))
         image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
-        processed_image_cpu = transforms.image_transform(image.copy())
+        processed_image_cpu = image_transform(image.copy())
         processed_image = processed_image_cpu.contiguous().to(args.device, non_blocking=True)
         fields = pifpaf.fields(torch.unsqueeze(processed_image, 0))[0]
         _, _, pifpaf_out = pifpaf.forward(image, processed_image_cpu, fields)

setup.py

@@ -27,7 +27,7 @@ setup(
     zip_safe=False,
 
     install_requires=[
-        'openpifpaf==0.8.0',
+        'openpifpaf',
         'tabulate',   # For evaluation
     ],
     extras_require={