Timeline for Fine-tuning a model from an existing checkpoint with TensorFlow-Slim
Current License: CC BY-SA 3.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 13, 2017 at 8:53 | comment | added | Shamane Siriwardhana | How to change the final layer classes ? The tutorial hasn't give much on that. | |
| Mar 12, 2017 at 13:26 | vote | accept | Alwyn Mathew | ||
| Mar 12, 2017 at 12:02 | comment | added | Icyblade | I'm sorry but I'm not a tf-slim user :-( | |
| Mar 12, 2017 at 11:51 | comment | added | Alwyn Mathew | Do you have any tips that will help me implement the dataset part myself? | |
| Mar 12, 2017 at 11:32 | comment | added | Icyblade |
flowers is actually a built-in dataset. Checkout github.com/tensorflow/models/blob/master/slim/datasets/…
|
|
| Mar 12, 2017 at 11:22 | comment | added | Alwyn Mathew | But the example in the question shows how to fine tune pretrained inception model on imagenet, with custom dataset flower. | |
| Mar 12, 2017 at 11:13 | comment | added | Icyblade |
After reading the source code, I found out train_image_classifier.py script didn't provides custom dataset support. Dataset is restricted here and here. If you insist fine-tuning with tf-slim, maybe you need to implement the dataset part yourself.
|
|
| Mar 12, 2017 at 11:01 | comment | added | Alwyn Mathew |
If I'm using a custom dataset, why should I give dataset_name? Is that train_dir works just like:file_writer = tf.summary.FileWriter('/path/to/logs', sess.graph) collecting event logs. How do you find these list of scopes of variables to be exclude or list of scopes to filter the set of variables to train without interrupting the model?
|
|
| Mar 12, 2017 at 9:49 | history | answered | Icyblade | CC BY-SA 3.0 |