crosstyan/RapidPoseTriangulation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Speed up preprocessing.

Browse Source
This commit is contained in:
Daniel
2024-11-29 16:19:06 +01:00
parent 93d4611a91
commit 1b5e0c44e3
4 changed files with 24 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
extras/easypose/detection.py
View File

@ -19,25 +19,26 @@ class RTMDet(BaseModel):
self.dy = 0
self.scale = 0
norm_mean = -1 * np.array([123.675, 116.28, 103.53])
norm_std = 1.0 / np.array([58.395, 57.12, 57.375])
self.norm_mean = np.reshape(norm_mean, (1, 1, 3)).astype(np.float32)
self.norm_std = np.reshape(norm_std, (1, 1, 3)).astype(np.float32)
def preprocess(self, image: np.ndarray):
th, tw = self.input_shape[2:]
tensor, self.dx, self.dy, self.scale = letterbox(
image, (tw, th), fill_value=114
)
tensor -= np.array((123.675, 116.28, 103.53))
tensor /= np.array((58.395, 57.12, 57.375))
tensor = tensor.astype(np.float32, copy=False)
tensor += self.norm_mean
tensor *= self.norm_std
tensor = tensor[..., ::-1]
tensor = (
np.expand_dims(tensor, axis=0).transpose((0, 3, 1, 2)).astype(np.float32)
)
tensor = np.expand_dims(tensor, axis=0).transpose((0, 3, 1, 2))
return tensor
def postprocess(self, tensor: List[np.ndarray]):
boxes = tensor[0]
classes = tensor[1]
boxes = np.squeeze(boxes, axis=0)
classes = np.squeeze(classes, axis=0)
classes = np.expand_dims(classes, axis=-1)
boxes = np.squeeze(tensor[0], axis=0)
classes = np.expand_dims(np.squeeze(tensor[1], axis=0), axis=-1)
boxes = np.concatenate([boxes, classes], axis=-1)
boxes = nms(boxes, self.iou_threshold, self.conf_threshold)

3
extras/easypose/pipeline.py
View File

@ -199,7 +199,8 @@ class TopDown:
# See: /mmpose/models/pose_estimators/topdown.py - add_pred_to_datasample()
th, tw = region.shape[:2]
bw, bh = [p.box[2] - p.box[0], p.box[3] - p.box[1]]
kp[:, :2] = kp[:, :2] / np.array([tw, th]) * np.array([bw, bh])
kp[:, :2] /= np.array([tw, th])
kp[:, :2] *= np.array([bw, bh])
kp[:, :2] += np.array([p.box[0] + bw / 2, p.box[1] + bh / 2])
kp[:, :2] -= 0.5 * np.array([bw, bh])

14
extras/easypose/pose.py
View File

@ -43,13 +43,17 @@ class SimCC(BaseModel):
self.dy = 0
self.scale = 0
norm_mean = -1 * np.array([123.675, 116.28, 103.53])
norm_std = 1.0 / np.array([58.395, 57.12, 57.375])
self.norm_mean = np.reshape(norm_mean, (1, 1, 3)).astype(np.float32)
self.norm_std = np.reshape(norm_std, (1, 1, 3)).astype(np.float32)
def preprocess(self, image: np.ndarray):
tensor, self.dx, self.dy, self.scale = image, 0, 0, 1
tensor -= np.array((123.675, 116.28, 103.53))
tensor /= np.array((58.395, 57.12, 57.375))
tensor = (
np.expand_dims(tensor, axis=0).transpose((0, 3, 1, 2)).astype(np.float32)
)
tensor = tensor.astype(np.float32, copy=False)
tensor += self.norm_mean
tensor *= self.norm_std
tensor = np.expand_dims(tensor, axis=0).transpose((0, 3, 1, 2))
return tensor
def postprocess(self, tensor: List[np.ndarray]):

7
extras/easypose/utils.py
View File

@ -11,13 +11,10 @@ def letterbox(img: np.ndarray, target_size: Sequence[int], fill_value: int = 128
scale = min(tw / w, th / h)
nw, nh = int(w * scale), int(h * scale)
resized_img = cv2.resize(img, (nw, nh))
dx, dy = (tw - nw) // 2, (th - nh) // 2
canvas = np.full((th, tw, img.shape[2]), fill_value, dtype=img.dtype)
dx, dy = (tw - nw) // 2, (th - nh) // 2
canvas[dy:dy + nh, dx:dx + nw, :] = resized_img
canvas[dy:dy + nh, dx:dx + nw, :] = cv2.resize(img, (nw, nh))
return canvas, dx, dy, scale