Nesne Algılama ve Kuşatan Kutular

import torch
def box_corner_to_center(boxes):
    (sol üst, sağ alt) konumundan (merkez, genişlik, yükseklik) konumuna dönüştür.
    x1, y1, x2, y2 = boxes[:, 0], boxes[:, 1], boxes[:, 2], boxes[:, 3]
    cx = (x1 + x2) / 2
    cy = (y1 + y2) / 2
    w = x2 - x1
    h = y2 - y1
    boxes = torch.stack((cx, cy, w, h), axis=-1)
    return boxes
def box_center_to_corner(boxes):
    (merkez, genişlik, yükseklik) konumundan (sol üst, sağ alt) konumuna dönüştür.
    cx, cy, w, h = boxes[:, 0], boxes[:, 1], boxes[:, 2], boxes[:, 3]
    x1 = cx - 0.5 * w
    y1 = cy - 0.5 * h
    x2 = cx + 0.5 * w
    y2 = cy + 0.5 * h
    boxes = torch.stack((x1, y1, x2, y2), axis=-1)
    return boxes
# Test verisi ile doğrulama
dog_bbox, cat_bbox = [60.0, 45.0, 378.0, 516.0], [400.0, 112.0, 655.0, 493.0]
boxes = torch.tensor([dog_bbox, cat_bbox])
print(box_center_to_corner(box_corner_to_center(boxes)) == boxes)
Kuşatan Kutuları Çizme

import matplotlib.pyplot as plt
import matplotlib.patches as patches
def bbox_to_rect(bbox, color):
    Kuşatan kutuyu matplotlib biçimine dönüştürün.
    return patches.Rectangle(
        (bbox[0], bbox[1]), bbox[2] - bbox[0], bbox[3] - bbox[1],
        linewidth=2, edgecolor=color, facecolor='none')
img = plt.imread('catdog.jpg')
fig, ax = plt.subplots(1)
ax.imshow(img)
dog_bbox = [60.0, 45.0, 378.0, 516.0]
cat_bbox = [400.0, 112.0, 655.0, 493.0]
ax.add_patch(bbox_to_rect(dog_bbox, 'blue'))
ax.add_patch(bbox_to_rect(cat_bbox, 'red'))
plt.show()
Çapa Kutuları Oluşturma
python
def multibox_prior(data, sizes, ratios):
    Her pikselde ortalanmış farklı şekillere sahip çapa kutuları oluşturun.
    in_height, in_width = data.shape[-2:]
    device, num_sizes, num_ratios = data.device, len(sizes), len(ratios)
    boxes_per_pixel = (num_sizes + num_ratios - 1)
    size_tensor = torch.tensor(sizes, device=device)
    ratio_tensor = torch.tensor(ratios, device=device)
    offset_h, offset_w = 0.5, 0.5
    steps_h = 1.0 / in_height
    steps_w = 1.0 / in_width
    center_h = (torch.arange(in_height, device=device) + offset_h) * steps_h
    center_w = (torch.arange(in_width, device=device) + offset_w) * steps_w
    shift_y, shift_x = torch.meshgrid(center_h, center_w)
    shift_y, shift_x = shift_y.reshape(-1), shift_x.reshape(-1)
    w = torch.cat((size_tensor * torch.sqrt(ratio_tensor[0]),
                   sizes[0] * torch.sqrt(ratio_tensor[1:]))) * in_height / in_width
    h = torch.cat((size_tensor / torch.sqrt(ratio_tensor[0]),
                   sizes[0] / torch.sqrt(ratio_tensor[1:])))
    anchor_manipulations = torch.stack((-w, -h, w, h)).T.repeat(
        in_height * in_width, 1) / 2
    out_grid = torch.stack([shift_x, shift_y, shift_x, shift_y],
                           dim=1).repeat_interleave(boxes_per_pixel, dim=0)
    output = out_grid + anchor_manipulations
    return output.unsqueeze(0)
img = plt.imread('catdog.jpg')
h, w = img.shape[:2]
X = torch.rand(size=(1, 3, h, w))  # Girdi verisi oluştur
Y = multibox_prior(X, sizes=[0.75, 0.5, 0.25], ratios=[1, 2, 0.5])
print(Y.shape)  # (1, çapa kutusu sayısı, 4)
Kesişim Birleşim Üzerinde (IoU) Hesaplama
python
def box_iou(boxes1, boxes2):
    İki çapa veya kuşatan kutu listesinde ikili IoU hesaplayın.
    box_area = lambda boxes: ((boxes[:, 2] - boxes[:, 0]) *
                              (boxes[:, 3] - boxes[:, 1]))
    areas1 = box_area(boxes1)
    areas2 = box_area(boxes2) 
    inter_upperlefts = torch.max(boxes1[:, None, :2], boxes2[:, :2])
    inter_lowerrights = torch.min(boxes1[:, None, 2:], boxes2[:, 2:])
    inters = (inter_lowerrights - inter_upperlefts).clamp(min=0)
    inter_areas = inters[:, :, 0] * inters[:, :, 1]
    union_areas = areas1[:, None] + areas2 - inter_areas   
    return inter_areas / union_areas
# Örnek verilerle test
boxes1 = torch.tensor([[0.1, 0.1, 0.2, 0.2], [0.2, 0.2, 0.3, 0.3]])
boxes2 = torch.tensor([[0.15, 0.15, 0.25, 0.25], [0.25, 0.25, 0.35, 0.35]])
print(box_iou(boxes1, boxes2))

def assign_anchor_to_bbox(ground_truth, anchors, device, iou_threshold=0.5):
    En yakın gerçek referans değeri kuşatan kutuları çapa kutularına atayın.
    num_anchors, num_gt_boxes = anchors.shape[0], ground_truth.shape[0]
    jaccard = box_iou(anchors, ground_truth)
    anchors_bbox_map = torch.full((num_anchors,), -1, dtype=torch.long, device=device) 
    max_ious, indices = torch.max(jaccard, dim=1)
    anc_i = torch.nonzero(max_ious >= iou_threshold).reshape(-1)
    box_j = indices[max_ious >= iou_threshold]
    anchors_bbox_map[anc_i] = box_j 
    for _ in range(num_gt_boxes):
        max_idx = torch.argmax(jaccard)
        box_idx = (max_idx % num_gt_boxes).long()
        anc_idx = (max_idx // num_gt_boxes).long()
        anchors_bbox_map[anc_idx] = box_idx
        jaccard[:, box_idx] = -1
        jaccard[anc_idx, :] = -1
    return anchors_bbox_map
# Örnek verilerle test
ground_truth = torch.tensor([[0.1, 0.1, 0.2, 0.2], [0.2, 0.2, 0.3, 0.3], [0.3, 0.3, 0.4, 0.4]])
anchors = torch.tensor([[0.1, 0.1, 0.2, 0.2], [0.2, 0.2, 0.3, 0.3], [0.3, 0.3, 0.4, 0.4], [0.4, 0.4, 0.5, 0.5]])
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
anchors_bbox_map = assign_anchor_to_bbox(ground_truth, anchors, device)
print(anchors_bbox_map)

def offset_boxes(anchors, assigned_bb, eps=1e-6):
    Çapa kutuları ile atanan kuşatan kutuları arasındaki offsetleri hesapla.
    c_anc = box_corner_to_center(anchors)
    c_assigned_bb = box_corner_to_center(assigned_bb)
    offset_xy = 10 * (c_assigned_bb[:, :2] - c_anc[:, :2]) / c_anc[:, 2:]
    offset_wh = 5 * torch.log(eps + c_assigned_bb[:, 2:] / c_anc[:, 2:])
    offset = torch.cat([offset_xy, offset_wh], axis=1)
    return offset
def multibox_target(anchors, labels):
    Etiketleri ve offsetleri hesapla.
    batch_size, anchors = labels.shape[0], anchors.squeeze(0)
    batch_offset, batch_mask, batch_cls_labels = [], [], []
    device, num_anchors = anchors.device, anchors.shape[0]
    for i in range(batch_size):
        label = labels[i, :, :]
        anchors_bbox_map = assign_anchor_to_bbox(label[:, 1:], anchors, device)
        bbox_mask = ((anchors_bbox_map >= 0).float().unsqueeze(-1)).repeat(1, 4)
        cls_labels = torch.zeros(num_anchors, dtype=torch.long, device=device)
        assigned_bb = torch.zeros((num_anchors, 4), dtype=torch.float32, device=device)
        indices_true = torch.nonzero(anchors_bbox_map >= 0)
        bb_idx = anchors_bbox_map[indices_true]
        assigned_bb[indices_true] = label[bb_idx, 1:]
        cls_labels[indices_true] = label[bb_idx, 0].long() + 1
        offset = offset_boxes(anchors, assigned_bb) * bbox_mask
        batch_offset.append(offset.reshape(-1))
        batch_mask.append(bbox_mask.reshape(-1))
        batch_cls_labels.append(cls_labels)
    return (torch.stack(batch_offset), torch.stack(batch_mask), torch.stack(batch_cls_labels))
# Örnek verilerle test
anchors = torch.tensor([[[0.1, 0.1, 0.2, 0.2], [0.2, 0.2, 0.3, 0.3], [0.3, 0.3, 0.4, 0.4], [0.4, 0.4, 0.5, 0.5]]])
labels = torch.tensor([[[0, 0.15, 0.15, 0.25, 0.25], [1, 0.25, 0.25, 0.35, 0.35]]])
print(multibox_target(anchors, labels))

def cls_loss(cls_preds, cls_labels):
    Sınıf tahminlerinin kaybını hesapla.
    return nn.CrossEntropyLoss(reduction='none')(cls_preds, cls_labels)
def bbox_loss(bbox_preds, bbox_labels, bbox_masks):
    Kuşatan kutu tahminlerinin kaybını hesapla.
    return (bbox_masks * F.smooth_l1_loss(bbox_preds, bbox_labels, reduction='none')).mean(dim=1)
# Örnek verilerle test
cls_preds = torch.tensor([[0.1, 0.2, 0.3, 0.4], [0.2, 0.3, 0.4, 0.5], [0.3, 0.4, 0.5, 0.6], [0.4, 0.5, 0.6, 0.7]])
cls_labels = torch.tensor([1, 0, 2, 3])
print(cls_loss(cls_preds, cls_labels))
bbox_preds = torch.tensor([[0.1, 0.2, 0.3, 0.4], [0.2, 0.3, 0.4, 0.5], [0.3, 0.4, 0.5, 0.6], [0.4, 0.5, 0.6, 0.7]])
bbox_labels = torch.tensor([[0.1, 0.2, 0.3, 0.4], [0.2, 0.3, 0.4, 0.5], [0.3, 0.4, 0.5, 0.6], [0.4, 0.5, 0.6, 0.7]])
bbox_masks = torch.tensor([[1.0, 1.0, 1.0, 1.0], [1.0, 1.0, 1.0, 1.0], [1.0, 1.0, 1.0, 1.0], [1.0, 1.0, 1.0, 1.0]])
print(bbox_loss(bbox_preds, bbox_labels, bbox_masks))