refactor: Cleaned codebase

README.md

@@ -66,6 +66,7 @@ conda install -c frgfm pylocron
 
 - Classification: [Res2Net](https://arxiv.org/abs/1904.01169) (based on the great [implementation](https://github.com/gasvn/Res2Net) from gasvn), [darknet24](https://pjreddie.com/media/files/papers/yolo_1.pdf), [darknet19](https://pjreddie.com/media/files/papers/YOLO9000.pdf), [darknet53](https://pjreddie.com/media/files/papers/YOLOv3.pdf).
 - Detection: [YOLOv1](https://pjreddie.com/media/files/papers/yolo_1.pdf), [YOLOv2](https://pjreddie.com/media/files/papers/YOLO9000.pdf)
+- Segmentation: [U-Net](https://arxiv.org/abs/1505.04597)
 
 ### ops
 

holocron/models/darknet.py

@@ -31,7 +31,6 @@
 
 
 class DarkBlockV1(nn.Sequential):
-
     def __init__(self, planes):
 
         layers = []
@@ -74,7 +73,6 @@ def forward(self, x):
 
 
 class DarknetV1(nn.Sequential):
-
     def __init__(self, layout, num_classes=10):
 
         super().__init__(OrderedDict([
@@ -86,7 +84,6 @@ def __init__(self, layout, num_classes=10):
 
 
 class DarkBlockV2(nn.Sequential):
-
     def __init__(self, in_planes, out_planes, nb_compressions=0):
 
         layers = [conv3x3(in_planes, out_planes),
@@ -104,7 +101,6 @@ def __init__(self, in_planes, out_planes, nb_compressions=0):
 
 
 class DarknetBodyV2(nn.Module):
-
     def __init__(self, layout, passthrough=False):
 
         super().__init__()
@@ -141,7 +137,6 @@ def forward(self, x):
 
 
 class DarknetV2(nn.Sequential):
-
     def __init__(self, layout, num_classes=10):
 
         super().__init__(OrderedDict([
@@ -153,14 +148,6 @@ def __init__(self, layout, num_classes=10):
 
 
 class DarkBlockV3(nn.Module):
-    """Implements a residual block of Darknet as described in
-    `"YOLOv3: An Incremental Improvement" <https://pjreddie.com/media/files/papers/YOLOv3.pdf>`_
-
-    Args:
-        planes (int): number of input/output channels
-        mid_planes (int): number of intermediate channels
-    """
-
     def __init__(self, planes, mid_planes):
 
         super().__init__()
@@ -192,7 +179,8 @@ def forward(self, x):
 
 class DarknetBodyV3(nn.Module):
 
-    def _make_layer(self, num_blocks, in_planes, out_planes=None):
+    @staticmethod
+    def _make_layer(num_blocks, in_planes, out_planes=None):
 
         layers = [DarkBlockV3(in_planes, in_planes // 2) for _ in range(num_blocks)]
         if isinstance(out_planes, int):
@@ -240,7 +228,6 @@ def forward(self, x):
 
 
 class DarknetV3(nn.Sequential):
-
     def __init__(self, layout, num_classes=10):
 
         super().__init__(OrderedDict([

holocron/models/detection/yolo.py

@@ -179,7 +179,6 @@ def post_process(b_coords, b_o, b_scores, rpn_nms_thresh=0.7, box_score_thresh=0
 
 
 class YOLOv1(_YOLO):
-
     def __init__(self, layout, num_classes=20, num_anchors=2, lambda_noobj=0.5, lambda_coords=5.):
 
         super().__init__()
@@ -285,7 +284,6 @@ def forward(self, x, gt_boxes=None, gt_labels=None):
 
 
 class YOLOv2(_YOLO):
-
     def __init__(self, layout, num_classes=20, anchors=None, lambda_noobj=0.5, lambda_coords=5.):
 
         super().__init__()

holocron/optim/lr_scheduler.py

@@ -115,7 +115,8 @@ def __init__(self,
 
         super(OneCycleScheduler, self).__init__(optimizer, last_epoch)
 
-    def _format_param(self, name, optimizer, param):
+    @staticmethod
+    def _format_param(name, optimizer, param):
         """Return correctly formatted lr/momentum for each param group."""
         if isinstance(param, (list, tuple)):
             if len(param) != len(optimizer.param_groups):

references/classification/train.py

@@ -20,7 +20,6 @@
 from torch import nn
 import torchvision
 from torchvision import transforms
-from torchvision.transforms import functional as F
 
 import holocron
 

references/detection/train.py

@@ -7,18 +7,14 @@
 
 import math
 import datetime
-import os
 import time
 from pathlib import Path
-from tqdm import tqdm
 from fastprogress import master_bar, progress_bar
 import matplotlib.pyplot as plt
 from matplotlib.patches import Rectangle
 
 import torch
 import torch.utils.data
-from torch import nn
-import torchvision
 from torchvision import transforms
 from torchvision.datasets import VOCDetection
 from torchvision.ops.boxes import box_iou
@@ -179,7 +175,7 @@ def plot_lr_finder(train_batch, model, data_loader, optimizer, device,
 def plot_samples(images, targets):
     # Unnormalize image
     nb_samples = 4
-    fig, axes = plt.subplots(1, nb_samples, figsize=(20, 5))
+    _, axes = plt.subplots(1, nb_samples, figsize=(20, 5))
     for idx in range(nb_samples):
         img = images[idx]
         img *= torch.tensor([0.229, 0.224, 0.225]).view(-1, 1, 1)