def fasterrcnn_resnet50_fpn（）实例测试

训练阶段

import torch
import torchvision
from torchvision.models.detection import FasterRCNN_ResNet50_FPN_Weights

model = torchvision.models.detection.fasterrcnn_resnet50_fpn(weights=FasterRCNN_ResNet50_FPN_Weights.DEFAULT)

# for train
# 以前一直有误区以为rand会生成4，或者3，这不是乱套了嘛，应该按顺序生成呀，实际上这是维数。浅显理解就是逐渐堆叠,先生成1200列，再生成600个这样的，再生成3个这样的。。。。。
images, boxes = torch.rand(4, 3, 600, 1200), torch.rand(4, 11, 4)
# 为了保证x2，y2一定大于x1，y1
boxes[:, :, 2:4] = boxes[:, :, 0:2] + boxes[:, :, 2:4]
# 前面参数是randint的最大最小值，（4，11）是生成的列表形状
labels = torch.randint(1, 91, (4, 11))

# 把每个图片给单独放到list里面，符合input的格式
images = list(image for image in images)

#targets是字典的列表，每个列表包括boxes和labels
targets = []
for i in range(len(images)):  # i = 0, 1, 2, 3
    d = {}
    d['boxes'] = boxes[i]
    d['labels'] = labels[i]
    targets.append(d)

output = model(images, targets)
print(output)
#
# # for inference
# model.eval()
#
# x = [torch.rand(3, 300, 400), torch.rand(3, 300, 400)]
# predictions = model(x)

结果：

{'loss_classifier': tensor(0.3633, grad_fn=<NllLossBackward0>), 'loss_box_reg': tensor(0.0262, grad_fn=<DivBackward0>), 'loss_objectness': tensor(1.9085, grad_fn=<BinaryCrossEntropyWithLogitsBackward0>), 'loss_rpn_box_reg': tensor(1.0729, grad_fn=<DivBackward0>)}

推理阶段

import torch
import torchvision
from torchvision.models.detection import FasterRCNN_ResNet50_FPN_Weights

model = torchvision.models.detection.fasterrcnn_resnet50_fpn(weights=FasterRCNN_ResNet50_FPN_Weights.DEFAULT)

# # for train
# images, boxes = torch.rand(4, 3, 600, 1200), torch.rand(4, 11, 4)
# boxes[:, :, 2:4] = boxes[:, :, 0:2] + boxes[:, :, 2:4]
# labels = torch.randint(1, 91, (4, 11))
#
# images = list(image for image in images)
# targets = []
# for i in range(len(images)):  # i = 0, 1, 2, 3
#     d = {}
#     d['boxes'] = boxes[i]
#     d['labels'] = labels[i]
#     targets.append(d)
#
# output = model(images, targets)
# print(output)
#
# for inference
model.eval()

x = [torch.rand(3, 300, 400), torch.rand(3, 300, 400)]
predictions = model(x)
print(predictions)

结果：
推测没有预测框的原因是数据是随机生成的

[{'boxes': tensor([], size=(0, 4), grad_fn=<StackBackward0>), 'labels': tensor([], dtype=torch.int64), 'scores': tensor([], grad_fn=<IndexBackward0>)}, 
{'boxes': tensor([], size=(0, 4), grad_fn=<StackBackward0>), 'labels': tensor([], dtype=torch.int64), 'scores': tensor([], grad_fn=<IndexBackward0>)}]