Some mixed improvements.
This commit is contained in:
Daniel
@ -37,14 +37,16 @@ class BaseModel(ABC):
|
||||
self.input_types = []
|
||||
for i in range(len(input_types)):
|
||||
input_type = input_types[i]
|
||||
if input_type == "tensor(float16)":
|
||||
if input_type == "tensor(float32)":
|
||||
itype = np.float32
|
||||
elif input_type == "tensor(float16)":
|
||||
itype = np.float16
|
||||
elif input_type == "tensor(uint8)":
|
||||
itype = np.uint8
|
||||
elif input_type == "tensor(int32)":
|
||||
itype = np.int32
|
||||
elif input_type == "tensor(uint8)":
|
||||
itype = np.uint8
|
||||
else:
|
||||
itype = np.float32
|
||||
raise ValueError("Undefined input type:", input_type)
|
||||
self.input_types.append(itype)
|
||||
|
||||
if warmup > 0:
|
||||
@ -59,6 +61,8 @@ class BaseModel(ABC):
|
||||
pass
|
||||
|
||||
def warmup(self, epoch: int):
|
||||
np.random.seed(42)
|
||||
|
||||
print("Running warmup for '{}' ...".format(self.__class__.__name__))
|
||||
for _ in tqdm(range(epoch)):
|
||||
inputs = {}
|
||||
@ -139,20 +143,30 @@ class LetterBox:
|
||||
def resize_image(self, image):
|
||||
paddings, _, new_size = self.calc_params(image.shape)
|
||||
|
||||
target_h, target_w = self.target_size
|
||||
canvas = np.full(
|
||||
(target_h, target_w, image.shape[2]),
|
||||
self.fill_value,
|
||||
dtype=image.dtype,
|
||||
)
|
||||
|
||||
# Resize the image
|
||||
new_w, new_h = new_size
|
||||
dx, dy = paddings[0], paddings[2]
|
||||
canvas[dy : dy + new_h, dx : dx + new_w, :] = cv2.resize(
|
||||
image, (new_w, new_h), interpolation=cv2.INTER_LINEAR
|
||||
resized_img = cv2.resize(
|
||||
image,
|
||||
(new_w, new_h),
|
||||
interpolation=cv2.INTER_LINEAR,
|
||||
)
|
||||
|
||||
return canvas
|
||||
# Optionally pad the image
|
||||
pad_left, pad_right, pad_top, pad_bottom = paddings
|
||||
if pad_left == 0 and pad_right == 0 and pad_top == 0 and pad_bottom == 0:
|
||||
final_img = resized_img
|
||||
else:
|
||||
final_img = cv2.copyMakeBorder(
|
||||
resized_img,
|
||||
pad_top,
|
||||
pad_bottom,
|
||||
pad_left,
|
||||
pad_right,
|
||||
borderType=cv2.BORDER_CONSTANT,
|
||||
value=[self.fill_value, self.fill_value, self.fill_value],
|
||||
)
|
||||
|
||||
return final_img
|
||||
|
||||
|
||||
# ==================================================================================================
|
||||
@ -211,6 +225,7 @@ class BoxCrop:
|
||||
new_end_y = min(ishape[0] - 1, end_y)
|
||||
new_box = [new_start_x, new_start_y, new_end_x, new_end_y]
|
||||
|
||||
# Calculate resized crop size
|
||||
bbox_w = new_box[2] - new_box[0]
|
||||
bbox_h = new_box[3] - new_box[1]
|
||||
scale = min(target_w / bbox_w, target_h / bbox_h)
|
||||
@ -250,22 +265,33 @@ class BoxCrop:
|
||||
def crop_resize_box(self, image, bbox):
|
||||
paddings, _, new_box, new_size = self.calc_params(image.shape, bbox)
|
||||
|
||||
image = image[new_box[1] : new_box[3], new_box[0] : new_box[2]]
|
||||
# Extract the bounding box
|
||||
cropped_img = image[new_box[1] : new_box[3], new_box[0] : new_box[2]]
|
||||
|
||||
th, tw = self.target_size
|
||||
canvas = np.full(
|
||||
(th, tw, image.shape[2]),
|
||||
self.fill_value,
|
||||
dtype=image.dtype,
|
||||
# Resize the image
|
||||
new_w, new_h = new_size
|
||||
resized_img = cv2.resize(
|
||||
cropped_img,
|
||||
(new_w, new_h),
|
||||
interpolation=cv2.INTER_LINEAR,
|
||||
)
|
||||
|
||||
nw, nh = new_size
|
||||
dx, dy = paddings[0], paddings[2]
|
||||
canvas[dy : dy + nh, dx : dx + nw, :] = cv2.resize(
|
||||
image, (nw, nh), interpolation=cv2.INTER_LINEAR
|
||||
)
|
||||
# Optionally pad the image
|
||||
pad_left, pad_right, pad_top, pad_bottom = paddings
|
||||
if pad_left == 0 and pad_right == 0 and pad_top == 0 and pad_bottom == 0:
|
||||
final_img = resized_img
|
||||
else:
|
||||
final_img = cv2.copyMakeBorder(
|
||||
resized_img,
|
||||
pad_top,
|
||||
pad_bottom,
|
||||
pad_left,
|
||||
pad_right,
|
||||
borderType=cv2.BORDER_CONSTANT,
|
||||
value=[self.fill_value, self.fill_value, self.fill_value],
|
||||
)
|
||||
|
||||
return canvas
|
||||
return final_img
|
||||
|
||||
|
||||
# ==================================================================================================
|
||||
@ -308,27 +334,17 @@ class RTMDet(BaseModel):
|
||||
boxes[:, 3] -= paddings[2]
|
||||
|
||||
boxes = np.maximum(boxes, 0)
|
||||
|
||||
th, tw = self.target_size
|
||||
pad_w = paddings[0] + paddings[1]
|
||||
pad_h = paddings[2] + paddings[3]
|
||||
max_w = tw - pad_w - 1
|
||||
max_h = th - pad_h - 1
|
||||
b0 = boxes[:, 0]
|
||||
b1 = boxes[:, 1]
|
||||
b2 = boxes[:, 2]
|
||||
b3 = boxes[:, 3]
|
||||
b0 = np.minimum(b0, max_w)
|
||||
b1 = np.minimum(b1, max_h)
|
||||
b2 = np.minimum(b2, max_w)
|
||||
b3 = np.minimum(b3, max_h)
|
||||
boxes[:, 0] = b0
|
||||
boxes[:, 1] = b1
|
||||
boxes[:, 2] = b2
|
||||
boxes[:, 3] = b3
|
||||
boxes[:, 0] = np.minimum(boxes[:, 0], max_w)
|
||||
boxes[:, 1] = np.minimum(boxes[:, 1], max_h)
|
||||
boxes[:, 2] = np.minimum(boxes[:, 2], max_w)
|
||||
boxes[:, 3] = np.minimum(boxes[:, 3], max_h)
|
||||
|
||||
boxes[:, 0:4] /= scale
|
||||
|
||||
return boxes
|
||||
|
||||
|
||||
@ -342,8 +358,6 @@ class RTMPose(BaseModel):
|
||||
self.boxcrop = BoxCrop(self.target_size, padding_scale=1.25, fill_value=0)
|
||||
|
||||
def preprocess(self, image: np.ndarray, bbox: np.ndarray):
|
||||
tensor = np.asarray(image).astype(self.input_types[0], copy=False)
|
||||
tensor = np.expand_dims(tensor, axis=0)
|
||||
bbox = np.asarray(bbox)[0:4]
|
||||
bbox += np.array([-0.5, -0.5, 0.5 - 1e-8, 0.5 - 1e-8])
|
||||
bbox = bbox.round().astype(np.int32)
|
||||
@ -368,12 +382,8 @@ class RTMPose(BaseModel):
|
||||
kp[:, 0:2] = np.maximum(kp[:, 0:2], 0)
|
||||
max_w = image.shape[1] - 1
|
||||
max_h = image.shape[0] - 1
|
||||
b0 = kp[:, 0]
|
||||
b1 = kp[:, 1]
|
||||
b0 = np.minimum(b0, max_w)
|
||||
b1 = np.minimum(b1, max_h)
|
||||
kp[:, 0] = b0
|
||||
kp[:, 1] = b1
|
||||
kp[:, 0] = np.minimum(kp[:, 0], max_w)
|
||||
kp[:, 1] = np.minimum(kp[:, 1], max_h)
|
||||
|
||||
return kp
|
||||
|
||||
|
||||
Reference in New Issue
Block a user