I am a beginner for tensorflow. I had just tried to fit a simple LeNet-5 for mnist data.
My training and test data are first in Numpy format. i.e., (60000, 28, 28). Then I set my model as below.
model_LeNet5 = Sequential([
layers.Conv2D(6, kernel_size=3, strides=1, input_shape=(28, 28, 1)),
layers.MaxPooling2D(pool_size=2,strides=2),
layers.ReLU(),
layers.Conv2D(16,kernel_size=3,strides=1),
layers.MaxPooling2D(pool_size=2,strides=2),
layers.ReLU(),
layers.Flatten(),
layers.Dense(120, activation='relu'),
layers.Dense(84, activation='relu'),
layers.Dense(10)
])
I could understand that I get success when I set input_shape as (28,28) or train_images.shape[1:], but I can not understand that input_shape = (28,28,1) is also worked (shown as code above).
It seems that there is an inconsistancy between the shape of data and setting of input size (i.e., [60000,28,28] vs [28,28,1]). Also the broadcast rule may not link [60000,28,28] with [28,28,1].
Thanks for anyone who will explain the mechanism of input_shape.
A single grayscale image can be represented using a two-dimensional (2D) NumPy array or a tensor. Since there is only one channel in a grayscale image, we don’t need an extra dimension to represent the color channel. The two dimensions represent the height and width of the image.
A batch of 3 grayscale images can be represented using a three-dimensional (3D) NumPy array or a tensor. Here, we need an extra dimension to represent the number of images.
For more information, check out this article on towardsdatascience.
Related
I am making a face mask detection project and I trained my model using ultralytics/yolov5.I saved the trained model as an onnx file, you can find the model file here model.onnx. Now I want you use this model.onnx with opencv to detect real time face mask. The input image size during training was 320*320. You can visualize this model using netron.
I have written this code to capture the image using webcam and pass it to model.onnx to predict my bounding boxes. The code is as follows:
def predict(img):
session = onnxruntime.InferenceSession(model_path)
input_name = session.get_inputs()[0].name
output_name = session.get_outputs()[0].name
img = img.reshape((1,3,320,320))
data = json.dumps({'data':img.tolist()})
data = np.array(json.loads(data)['data']).astype('float32')
result = session.run([output_name],{input_name:data})
result = np.array(result)
print(result.shape)
The output of result.shape is (1, 1, 3, 40, 40, 85)
Can anyone help me in interpreting this shape and how can i use this result array to predict my class, bounding box and confidence.
I've never worked with a pure yolov5 model, but here's the output format for yolov5s. It looks like it should be similar.
ouput tensor structure (yolov5s):
output_tensor[a, b, c, d]
a -> image index (If you're input is a batch of images, this tells you which image's output you're looking at. If your input is just one image, leave this as 0.)
b -> index of image in batch
c -> information about bounding box
0, 1 -> x and y coordinate of bounding box center
2, 3 -> width and height of bounding box
4 -> bounding box confidence
5 - 85 -> single class confidences
d -> index of proposed bounding boxes
I am trying to use demo.py in nkolot
/
GraphCMR | GitHub. I am interested in obtaining joints from the inferred SMPL image and visualize it similar to described in README of this project: gulvarol
/
smplpytorch | GitHub.
I also posted the issue here: https://github.com/nkolot/GraphCMR/issues/36.
What I tried that didn't work.
I changed https://github.com/nkolot/GraphCMR/blob/4e57dca4e9da305df99383ea6312e2b3de78c321/demo.py#L118 to
pred_vertices, pred_vertices_smpl, pred_camera, smpl_pose, smpl_shape = model(...) to get smpl_pose (of shape torch.Size([1, 24, 3, 3])). Then I just flattened it by doing smpl_pose.cpu().data.numpy()[:, :, :, -1].flatten('C').reshape(1, -1) and used the resulting (1, 72) pose params as input in pose_params variable of smplpytorch demo.
The resulting visualization doesn't look correct to me. Is this the right approach? Perhaps there is an easier way to do what I am doing.
How to get 3d joints from demo.py and visualize it | nkolot
/
GraphCMR
The problem is that
smpl_pose (of shape torch.Size([1, 24, 3, 3]))
is the SMPL pose parameters expressed as a rotation matrix.
You need to make a transformation from rotation matrix to axis-angle representation which is (72,1). You can use Rodrigues formula to do it, as claimed in the paper:
Get more information from the paper:
SMPL: A Skinned Multi-Person Linear Model
I am interested in generating weighted, directed random graphs with node constraints. Is there a graph generator in R or Python that is customizable? The only one I am aware of is igraph's erdos.renyi.game() but I am unsure if one can customize it.
Edit: the customizations I want to make are 1) drawing a weighted graph and 2) constraining some nodes from drawing edges.
In igraph python, you can use link the Erdos_Renyi class.
For constraining some nodes from drawing edges, this is controlled by the p value.
Erdos_Renyi(n, p, m, directed=False, loops=False) #these are the defaults
Example:
from igraph import *
g = Graph.Erdos_Renyi(10,0.1,directed=True)
plot(g)
By setting the p=0.1 you can see that some nodes do not have edges.
For the weights you can do something like:
g.ecount() # to find the number of edges
g.es["weights"] = range(1, g.ecount())
g.es["label"] = weights
plot(g)
Result:
I would like to use either dxfwrite or ezdxf to create text along (WCS) y direction, and with height in the (WCS) z direction.
Using autocad, I have done this by setting UCS and entering text.
How can I do in dxfwrite or ezdxf (or any other python friendly library)?
dxf.ucs('textucs',xaxis=(0.,1.,0),yaxis=(0.,0.,1.))
lab = dxf.mtext('hello',np.array([0.,0.,.5]),layer='mylay',height=0.3)
doesn't work, presumably because I have only created UCS, and am not using it.
Defining an UCS does nothing, dxfwrite/ezdxf are not CAD applications.
This example uses ezdxf to write a text in the YZ-plane:
import ezdxf
dwg = ezdxf.new('ac1015')
modelspace = dwg.modelspace()
modelspace.add_mtext("This is a text in the YZ-plane",
dxfattribs={
'width': 12, # reference rectangle width
'text_direction': (0, 1, 0), # write in y direction
'extrusion': (1, 0, 0) # normal vector of the text plane
})
dwg.saveas('mtext_in_yz_plane.dxf')
mtext in dxfwrite is just a bunch of TEXT entities, because the MTEXT entity requires DXF13 or later.
I have plots of 3-axis accelerometer time-series data (t,x,y,z) in separate subplots I'd like to zoom together. That is, when I use the "Zoom to Rectangle" tool on one plot, when I release the mouse all 3 plots zoom together.
Previously, I simply plotted all 3 axes on a single plot using different colors. But this is useful only with small amounts of data: I have over 2 million data points, so the last axis plotted obscures the other two. Hence the need for separate subplots.
I know I can capture matplotlib/pyplot mouse events (http://matplotlib.sourceforge.net/users/event_handling.html), and I know I can catch other events (http://matplotlib.sourceforge.net/api/backend_bases_api.html#matplotlib.backend_bases.ResizeEvent), but I don't know how to tell what zoom has been requested on any one subplot, and how to replicate it on the other two subplots.
I suspect I have the all the pieces, and need only that one last precious clue...
-BobC
The easiest way to do this is by using the sharex and/or sharey keywords when creating the axes:
from matplotlib import pyplot as plt
ax1 = plt.subplot(2,1,1)
ax1.plot(...)
ax2 = plt.subplot(2,1,2, sharex=ax1)
ax2.plot(...)
You can also do this with plt.subplots, if that's your style.
fig, ax = plt.subplots(3, 1, sharex=True, sharey=True)
Interactively this works on separate axes
for ax in fig.axes:
ax.set_xlim(0, 50)
fig.draw()