cuda toolkit archive: https://developer.nvidia.com/cuda-toolkit-archive
| Command | Description |
|---|---|
keypoints = outputs["instances"].pred_keypoints.to("cpu").detach().numpy() |
pytorch tensor to numpy |
t = torch.tensor([1, 2, 3], [4, 5, 6]) |
pytorch create a tensor (defualt type will be float32) |
t = torch.tensor([1, 2, 3], [4, 5, 6], dtype=torch.float64) |
pytorch create a tensor (specify type) |
t.shape |
pytorch tensor sahpe |
t.ndim |
pytorch tensor n dimention |
t.dtype |
pytorch tensor data types inside (e.g torch.float32) |
t.dot(t2) |
pytorch dot product |
assumeing
array_a = np.array([1,2,3,4,5])
| Command | Description |
|---|---|
array_b = np.where(array_a >= 3, -1, 1) |
array_b = [ 1 1 -1 -1 -1] |
array_b = np.where(array_a >= 3, array_a*10, array_a) |
array_b = [ 1 2 30 40 50] |
a = np.empty((0,3), int) |
Create a new numpy for appending or stacking |
img = np.zeros([100,100,3],dtype=np.uint8) then img.fill(255) # or img[:] = 255 |
create an empty black or white image |
id_x = np.where((points[:,0] > x_min) & (points[:,0] < x_max)) |
Get the id of desired part by filtering |
id_x = points[:,0] > x_min) & (points[:,0] < x_max |
Get boolean array to filter data |
np.linspace(min_value, max_value, num=int((max-min)/dist),endpoint=True) |
Create data or points between two number |
np.min(points,axis =0),np.max(points,axis =0),np.mean(points,axis =0) |
Get Min, Max, Mean of points |
points3D = np.vstack((f.x, f.y, f.z)).transpose() |
From las files to numpy in n*3 format |
np.linalg.norm(ppts2d - np.array([x, y]), axis=1) |
Calc distance of a vec elements to a point |
diff_to_min = ppts2d - np.array([x, y]) |
Calc difference of a vec elements to a point (signed) |
filter_axe = np.all(diff_to_min > 0, axis=1) |
find 2d points bigger than desired values in both x, y |
img_w = image.shape[0]
img_H = image.shape[1]
files = [line.strip().split() for line in open(os.path.join(self.label_dir, label_file))]
bboxes = [[int((float(item[1]) - (float(item[3]) / 2)) * img_w), # x_min
int((float(item[2]) - (float(item[4]) / 2)) * img_H), # y_min
int((float(item[1]) + (float(item[3]) / 2)) * img_w), # x_max
int((float(item[2]) + (float(item[4]) / 2)) * img_H)] # y_max
for item in files]
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent