583 questions
Best practices
0
votes
3
replies
48
views
How to detect mislabeled images in a directory based image classification dataset?
What is the best way to find mislabeled images in a dataset where labels are based on folders?
Each class is already separated into its own directory, but some images inside are clearly in the wrong ...
Tooling
0
votes
1
replies
297
views
Looking for a working TensorFlow Lite (.tflite) NSFW detection model for Android
I specifically need:
-A real .tflite file (not .onnx, not .pb, not .pth, not .mlmodel)
-Image classifier model (NOT segmenter or detector)
-Preferably with 5 classes or at least 2: drawings, hentai, ...
1
vote
0
answers
139
views
using Google LiteRT from a Tensorflow Lite model
data = tf.keras.utils.image_dataset_from_directory('snails', image_size=(256,256), shuffle=True)
class_names = data.class_names
num_classes = len(class_names)
print("Classes:", class_names)
...
-1
votes
1
answer
65
views
Get image paths from tfds food101
I'm working on food101 tensorflow dataset and want to know the most wrong predictions of my efficientnet model, for that purpose I'd need to get image paths of test data, but I don't know how I can ...
0
votes
1
answer
164
views
tflite-runtime ValueError: Didn't find op for builtin opcode 'FULLY_CONNECTED' version '12'. An older version of this builtin might be supported
I have trained and quantised my models on tensorflow==2.19.0 instead of a previous version and now am facing troubles in using the quantised models for inference on my Raspberry Pi (Raspbian GNU/Linux ...
- 45
0
votes
0
answers
23
views
Validation Accuracy stuck between 0.5-0.6
I am currently working on a custom VGG model for Breast Cancer Classification using mammography images. It has excellent accuracy for training images but the validation accuracy is stuck at 0.5-0.6. ...
1
vote
1
answer
193
views
DeepFace Analyze - Test scripts for batched image processing
I am having trouble getting image analysis batch processing to work. I would like to send a list of images to DeepFace to get a list of race, gender, and age predictions. I am calling test functions ...
- 13
0
votes
0
answers
32
views
Finetuning MobilenetV3S on Google Colab
Last year I finetuned MobileNetV3S on google colab [tensorflow version 2.15.0] gave me 95% (test) accuracy and loss below 0.25. However, running the same notebook now (2025-05) [tensorflow verion 2.18....
- 1
3
votes
0
answers
70
views
How to calculate FPS in image classification?
Please, I would appreciate some help and will be grateful for any help.
I have some doubts about how to calculate the frame rate (FPS) in image classification for training an testing routines. I've ...
- 31
2
votes
2
answers
3k
views
KeyError: 'Exception encountered when calling Functional.call()
I am trying to construct a model that accepts two inputs. Within this model, I've designed three blocks: two correspond to the structure of the first blocks of EfficientNetB0 up to block 4 (A), while ...
0
votes
2
answers
2k
views
Exception encountered: Unrecognized keyword arguments passed to Conv2D: {'batch_input_shape': [None, 135, 135, 3]}
I have some models that were trained with an older version of Tensorflow, Keras and Python. Now I updated to Tensorflow 2.16.1, Keras 3.1.1 and Python 3.12.2. Now I cannot even load these models that ...
0
votes
1
answer
228
views
How to load a Pre Trained VGG16 model using TensorFlow and generate a confusion matrix
I have trained a VGG16 model (from tensorflow.keras.applications) on 15 categories of images on 100 epochs. After training, I saved the model as a file named 'best_model.h5', but unfortunately, I ...
- 1
0
votes
1
answer
885
views
How to implement Grad-CAM to view activation map/heat maps on TensorFlow ResNet152V2 for image Classification
Hello I'm doing a small project on TensorFlow image classification using ResNet152V2.
I wrote a Train-Predict.py script that's able to train a trained_weights.hdf5 file to successfully predict images ...
- 45
-1
votes
1
answer
53
views
Convolutional Neural Network is getting unexpectedly high validation accuracy, what could be causing this?
I am building/optimizing a CNN for classifcation of cars from this dataset.
My baseline model is getting suprisingly high accuracy from a very simple model architecture, I am concerned there is some ...
0
votes
1
answer
312
views
Not able to load a Vision Transformer model post training
So I am using the example of the vision transformer model for image classification provided on the Keras website. The only difference is I have added a line to save the model once it is done training ...
- 3