machine-learningdeep-learningneural-networkcomputer-visionconv-neural-network
How to Handle Data of Different Dimensions in CNN
I'm very beginner to Machine Learning.
Working on a Character Recognition Model using CNN.I have image datasets of different sizes (mainly of two size,64X64 and 192X64), training data set is prepared by reshaping 64X64 to 32X32. This data sets works very well.
On other hand I too have 192X64 images on reshaping them into 32X32, the pixel data were broken and distorted.
Any guidance on how to handle these data would be very helpful.
Thanks in advance.
I tried reshaping the 32X32 as 192X64, I know it is way too much.Got Memory Error
Solution
There are three potential ways you can go:
- Can you crop 64x64 images from your 192x64 images without loosing too much information from other parts of the image? In this way your images do not get distorted by down-scaling.
- If cropping is not possible, make sure that your model trains on distorted (192x64 to 64x64 downscaled) data. If your model trains on these samples, there also should not be a problem during inference.
- You could be padding your image from 192x64 to 192x192, and then downsample. The sample will not be distorted, however there will be large areas with approximated / padded pixel values (definitely this kind of data also needs to be present during training if you choose to go this way).
- The “Forward/Backward Passage Size” is too large for the pytorch model (Yolov3)
- How many images(minimum) should be there in each classes for training YOLO?
- Why do neural networks work so well?
- Will larger batch size make computation time less in machine learning?
- Why KL divergence is negative in Pytorch?
- Creating a voice identification system using machine learning
- Forward pass with all samples
- fit method in sklearn
- Calibrating Probabilities in lightgbm or XGBoost
- Implementation of F1-score, IOU and Dice Score
- Is it ok to have the training history very similar to the validation history?
- How to understand Shapley value for binary classification problem?
- Stochastic Gradient Descent for Logistic Regression always returns a cost of Inf and weight vector never gets any closer
- Data format for Libsvm SVR training in Matlab
- Why shouldn't we use multiple activation functions in the same layer?
- How to implement a butterworth filter
- weka java api stringtovector exception
- Predict training data in sklearn
- Text classification with weka
- How can i apply feature reduction methods in Weka?
- 'super' object has no attribute '__sklearn_tags__'
- lightgbm.cv: cvbooster.best_iteration always returns -1
- Wrong detection from yolov5 model
- torchrl: Using SyncDataCollector with a custom pytorch dqn
- OCR Preprocessing for Oman License Plates - Issues with Alphabet Recognition
- Training difference between LightGBM API and Sklearn API
- Why can't I wrap LGBM?
- Why RAG is slower than LLM?
- Azure AutoML Image Classification Job
- What's the best way to use a sklearn feature selector in a grid search, to evaluate the usefulness of all features?