完善部署并训练5个新发型 + 换发型集成文档

部署修复： - torch.load 增加 weights_only=False patch，兼容 PyTorch 2.6+ 加载旧权重 - OSS 改为懒加载，本地用 output_format=base64 无需配凭证即可启动 - 补全被 gitignore 误排除的必需代码：core/models/layers/data、models/layers/data、keypoints/lib - webui 训练命令 --xformers 改 --sdpa（修复 xformers 无 CUDA 支持报错）功能调整： - hair_grow_service 端口改 8899、preview 路由修复（send_file） - list_hairstyles 增加发型白名单，测试页只展示当前5个发型新增脚本： - train_lora_parallel.py：直接调 kohya 并行训练 LoRA（绕过 photo_service 串行限制） - train_hairstyles_parallel.py / train_batch_stepC.py：批量训练辅助脚本 - scripts/sync_data_to_server.sh：大文件断点续传到云服务器文档： - docs/换发型集成文档.md：换发型完整流程、服务架构、资源依赖、训练方法、集成步骤
2026-07-09 20:45:08 +08:00
parent aaa1bf159f
commit ce445b64a6
42 changed files with 10553 additions and 34 deletions
@@ -0,0 +1,164 @@
+# ------------------------------------------------------------------------------
+# Copyright (c) Microsoft
+# Licensed under the MIT License.
+# Modified from py-faster-rcnn (https://github.com/rbgirshick/py-faster-rcnn)
+# ------------------------------------------------------------------------------
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import numpy as np
+
+def py_nms_wrapper(thresh):
+    def _nms(dets):
+        return nms(dets, thresh)
+    return _nms
+
+
+def nms(dets, thresh):
+    """
+    greedily select boxes with high confidence and overlap with current maximum <= thresh
+    rule out overlap >= thresh
+    :param dets: [[x1, y1, x2, y2 score]]
+    :param thresh: retain overlap < thresh
+    :return: indexes to keep
+    """
+    if dets.shape[0] == 0:
+        return []
+
+    x1 = dets[:, 0]
+    y1 = dets[:, 1]
+    x2 = dets[:, 2]
+    y2 = dets[:, 3]
+    scores = dets[:, 4]
+
+    areas = (x2 - x1 + 1) * (y2 - y1 + 1)
+    order = scores.argsort()[::-1]
+
+    keep = []
+    while order.size > 0:
+        i = order[0]
+        keep.append(i)
+        xx1 = np.maximum(x1[i], x1[order[1:]])
+        yy1 = np.maximum(y1[i], y1[order[1:]])
+        xx2 = np.minimum(x2[i], x2[order[1:]])
+        yy2 = np.minimum(y2[i], y2[order[1:]])
+
+        w = np.maximum(0.0, xx2 - xx1 + 1)
+        h = np.maximum(0.0, yy2 - yy1 + 1)
+        inter = w * h
+        ovr = inter / (areas[i] + areas[order[1:]] - inter)
+
+        inds = np.where(ovr <= thresh)[0]
+        order = order[inds + 1]
+
+    return keep
+
+
+def oks_iou(g, d, a_g, a_d, sigmas=None, in_vis_thre=None):
+    if not isinstance(sigmas, np.ndarray):
+        sigmas = np.array([.26, .25, .25, .35, .35, .79, .79, .72, .72, .62, .62, 1.07, 1.07, .87, .87, .89, .89]) / 10.0
+    vars = (sigmas * 2) ** 2
+    xg = g[0::3]
+    yg = g[1::3]
+    vg = g[2::3]
+    ious = np.zeros((d.shape[0]))
+    for n_d in range(0, d.shape[0]):
+        xd = d[n_d, 0::3]
+        yd = d[n_d, 1::3]
+        vd = d[n_d, 2::3]
+        dx = xd - xg
+        dy = yd - yg
+        e = (dx ** 2 + dy ** 2) / vars / ((a_g + a_d[n_d]) / 2 + np.spacing(1)) / 2
+        if in_vis_thre is not None:
+            ind = list(vg > in_vis_thre) and list(vd > in_vis_thre)
+            e = e[ind]
+        ious[n_d] = np.sum(np.exp(-e)) / e.shape[0] if e.shape[0] != 0 else 0.0
+    return ious
+
+
+def oks_nms(kpts_db, thresh, sigmas=None, in_vis_thre=None):
+    """
+    greedily select boxes with high confidence and overlap with current maximum <= thresh
+    rule out overlap >= thresh, overlap = oks
+    :param kpts_db
+    :param thresh: retain overlap < thresh
+    :return: indexes to keep
+    """
+    if len(kpts_db) == 0:
+        return []
+
+    scores = np.array([kpts_db[i]['score'] for i in range(len(kpts_db))])
+    kpts = np.array([kpts_db[i]['keypoints'].flatten() for i in range(len(kpts_db))])
+    areas = np.array([kpts_db[i]['area'] for i in range(len(kpts_db))])
+
+    order = scores.argsort()[::-1]
+
+    keep = []
+    while order.size > 0:
+        i = order[0]
+        keep.append(i)
+
+        oks_ovr = oks_iou(kpts[i], kpts[order[1:]], areas[i], areas[order[1:]], sigmas, in_vis_thre)
+
+        inds = np.where(oks_ovr <= thresh)[0]
+        order = order[inds + 1]
+
+    return keep
+
+
+def rescore(overlap, scores, thresh, type='gaussian'):
+    assert overlap.shape[0] == scores.shape[0]
+    if type == 'linear':
+        inds = np.where(overlap >= thresh)[0]
+        scores[inds] = scores[inds] * (1 - overlap[inds])
+    else:
+        scores = scores * np.exp(- overlap**2 / thresh)
+
+    return scores
+
+
+def soft_oks_nms(kpts_db, thresh, sigmas=None, in_vis_thre=None):
+    """
+    greedily select boxes with high confidence and overlap with current maximum <= thresh
+    rule out overlap >= thresh, overlap = oks
+    :param kpts_db
+    :param thresh: retain overlap < thresh
+    :return: indexes to keep
+    """
+    if len(kpts_db) == 0:
+        return []
+
+    scores = np.array([kpts_db[i]['score'] for i in range(len(kpts_db))])
+    kpts = np.array([kpts_db[i]['keypoints'].flatten() for i in range(len(kpts_db))])
+    areas = np.array([kpts_db[i]['area'] for i in range(len(kpts_db))])
+
+    order = scores.argsort()[::-1]
+    scores = scores[order]
+
+    # max_dets = order.size
+    max_dets = 20
+    keep = np.zeros(max_dets, dtype=np.intp)
+    keep_cnt = 0
+    while order.size > 0 and keep_cnt < max_dets:
+        i = order[0]
+
+        oks_ovr = oks_iou(kpts[i], kpts[order[1:]], areas[i], areas[order[1:]], sigmas, in_vis_thre)
+
+        order = order[1:]
+        scores = rescore(oks_ovr, scores[1:], thresh)
+
+        tmp = scores.argsort()[::-1]
+        order = order[tmp]
+        scores = scores[tmp]
+
+        keep[keep_cnt] = i
+        keep_cnt += 1
+
+    keep = keep[:keep_cnt]
+
+    return keep
+    # kpts_db = kpts_db[:keep_cnt]
+
+    # return kpts_db