Skip to main content

All Questions

Filter by
Sorted by
Tagged with
0 votes
0 answers
28 views

How do I solve the problem of displaying accuracy parameter in CNN?

I'm facing a problem while computing accuracy parameter and sketching it in a graph. I was expecting a drawn accuracy and loss graphs. In line: model.compile(optimizer='adam', loss='...
Ziad Mostafa's user avatar
1 vote
0 answers
377 views

Is there a way to change the backbone of Detectron2 keypoint detection model with custom object detection model?

I am working on a task where I have to detect people and do activity recognition. I am using **Detectron2 **Keypoint detection model.There are many False negatives with the detection. So my solution ...
sp0800's user avatar
  • 11
0 votes
1 answer
5k views

Detectron2 difference between SOLVER.IMS_PER_BATCH and MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE in cfg file

In detectron2 there seems to be two settings related to batch size in the config file. They are SOLVER.IMS_PER_BATCH and MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE I have seen both being used as batch size ...
devman3211's user avatar
1 vote
0 answers
133 views

visual embeddings with detectron

I'm trying to extract visual embeddings from the detectron model with the help of this colab notebook. I want the shape of vector after passing through fc1 to be (1000, 2048), but somehow this model ...
lazytux's user avatar
  • 307
0 votes
1 answer
93 views

One box object detection

I am using a faster rcnn model to predict one object in an image. There can only be one object in each image. Is it possible to force Faster Rcnn to train and predict as if it should only find one ...
Parash Dejmar's user avatar
1 vote
0 answers
479 views

Faster RCNN - Number of epochs impact inference time

I have trained two Faster RCNN models with the exact same parameters with the exception of: Model 1: 150 epochs Model 2: 5 epochs Once the models have been trained, I load them in and then I make ...
RaLa's user avatar
  • 21
1 vote
1 answer
622 views

Detecting object borders PRECISELY

I need to detect object (coin) in a photo, which is to draw a rectangle around it. I tried Mark-RCNN, Retina, Yolo, SSD - same result. The rectangles are not perfect. If you look at examples available ...
Steve Brown's user avatar
2 votes
0 answers
187 views

Symbol detection using Faster RCNN

I'm using a Faster RCNN network to perform object/symbol detection and I'm facing 2 major issues. The bounding box of the detected symbols is not tight enough. For example, in many cases, only 50%-70%...
Abhishek Sudhaker's user avatar
8 votes
3 answers
10k views

TensorFlow 2 Mask-RCNN? [closed]

I can't seem to find a reliable version of Mask-RCNN for TensorFlow 2. The matterport mask-rcnn (https://github.com/matterport/Mask_RCNN) has depreciated Tensorflow 1 code. Does anyone know of ...
myelin's user avatar
  • 99
2 votes
1 answer
2k views

Save the best model trained on Faster RCNN (COCO dataset) with Pytorch avoiding to "overfitting"

I am training a Faster RCNN neural network on COCO dataset with Pytorch. I have followed next tutorial: https://pytorch.org/tutorials/intermediate/torchvision_tutorial.html The training results are ...
Jose David's user avatar
1 vote
0 answers
511 views

prediction array is empty Mask RCNN

I am working on Mask RCNN and my prediction array is empty. I think it has to do with memory. This is what my code where I call for the predictions: model.eval() with torch.no_grad(): prediction = ...
E.W's user avatar
  • 11
0 votes
1 answer
42 views

Expected a dimension of size 5 but got array with shape (10000, 64, 173, 1) (CNN + LSTM on GTZAN dataset)

I am trying to implement a CNN + LSTM architecture on GTZAN dataset. I will explain my training and validation sets below: X_train.shape #(10000, 64, 173, 1) Y_train.shape #(10000, 10, 1) X_valid....
Esrat Maria's user avatar
5 votes
3 answers
10k views

Resnet-18 as backbone in Faster R-CNN

I code with pytorch and I want to use resnet-18 as backbone of Faster R-RCNN. When I print structure of resnet18, this is the output: >>import torch >>import torchvision >>import ...
CVDE's user avatar
  • 440
0 votes
1 answer
1k views

Reducing the Spatial Dimensions of a Neural Network Feature Map

Given a feature map of dimensionality MxNxC (for example, the output of a predicted Region of Interest from a Faster-RCNN), how would one reduce the spatial dimensions to be 1x1xC? I.e. reduce the ...
Miles's user avatar
  • 2,527
2 votes
0 answers
661 views

How Region proposal network works in Faster R-CNN?

I try to understand how anchor boxes coordinate generated from feature map, and i have some questions about this process. 1-) From above image, feature map size is N x M x C and sliding windows is ...
Bedrick Kiq's user avatar
0 votes
2 answers
1k views

What is the difference between test and validation specifically in Mask-R-CNN?

I have my own image dataset and use Mask-R-CNN for training. There you divide your dataset into train, valivation and test. I want to know the difference between validation and test. I know that ...
Mob's user avatar
  • 409
-3 votes
1 answer
387 views

choosing the best Pooling Type for my CNN

I could find max pooling is the most used and preferred type when it comes to Pooling, whatever the image data or the features i need to extract which is sound so ridicules to me for example i'm ...
moheb yossif's user avatar