我尝试使用以下方法保存和加载模型:
所有键都已映射,但输出中没有预测
#1
from detectron2.modeling import build_model
model = build_model(cfg)
torch.save(model.state_dict(), 'checkpoint.pth')
model.load_state_dict(torch.load(checkpoint_path,map_location='cpu'))
我也尝试使用官方文档来做,但无法理解输入格式部分from detectron2.checkpoint import DetectionCheckpointer
DetectionCheckpointer(model).load(file_path_or_url) # load a file, usually from cfg.MODEL.WEIGHTS
checkpointer = DetectionCheckpointer(model, save_dir="output")
checkpointer.save("model_999") # save to output/model_999.pth
最佳答案
cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file('COCO-Detection/faster_rcnn_R_101_FPN_3x.yaml'))
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5 # Set threshold for this model
cfg.MODEL.WEIGHTS = '/content/model_final.pth' # Set path model .pth
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 1
predictor = DefaultPredictor(cfg)
我加载自定义模型的代码有效。
关于deep-learning - 如何在 Detectron2 中保存和加载自定义数据集的模型?,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/63166152/