maskrcnn_benchmark理解记录——modeling\roi_heads\roi_heads.py

最新推荐文章于 2024-11-29 19:58:49 发布

业精于勤荒于嬉-行成于思而毁于随

最新推荐文章于 2024-11-29 19:58:49 发布

阅读量1.5k

点赞数

分类专栏： maskrcnn理解记录

本文链接：https://blog.csdn.net/m0_37644085/article/details/88699838

版权

本文档详细记录了maskrcnn_benchmark中modeling/roi_heads/roi_heads.py模块的理解，该模块包含检测和分割两个关键分支。通过对build_roi_keypoint_head的分析，深入探讨关键点检测头的实现细节。

摘要由CSDN通过智能技术生成

占坑

由两分支组成：检测分支和分割分支组成;

1)    roi_heads.append(("keypoint", build_roi_keypoint_head(cfg, in_channels)))
2)    roi_heads = CombinedROIHeads(cfg, roi_heads)

1)build_roi_keypoint_head可查看modeling\roi_heads\keypoint_head\keypoint_head.py

# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
import torch

from .box_head.box_head import build_roi_box_head
from .mask_head.mask_head import build_roi_mask_head
from .keypoint_head.keypoint_head import build_roi_keypoint_head


class CombinedROIHeads(torch.nn.ModuleDict):
    """
    Combines a set of individual heads (for box prediction or masks) into a single
    head.
    """

    def __init__(self, cfg, heads):
        super(CombinedROIHeads, self).__init__(heads)
        self.cfg = cfg.clone()
        if cfg.MO