Practical-Deep-Learning-Book

Practical-Deep-Learning-Book copied to clipboard

chapter-4/2-similarity-search-level-1.ipynb Broken float division

4

``` def plot_images(filenames, distances): images = [] for filename in filenames: images.append(mpimg.imread(filename)) plt.figure(figsize=(20, 10)) columns = 4 for i, image in enumerate(images): ax = plt.subplot(len(images) / columns + 1, columns,...

gitgithan

3 changes in this notebook to fix #163 #164 1. Quoting paths to prevent zsh shells erroring with `zsh:1: no matches found: https://drive.google.com/uc?id=137RyRjvTBkBiIfeYBNZBtViDHQ6_Ewsp` because it treats ? as glob character....

gitgithan

chapter-4/4-improving-accuracy-with-fine-tuning.ipynb does not contain the final finetuning section that is in book

1

There is a final finetuning throwing away dense and dropout layers that was expected according to the book but is missing in the notebook

gitgithan

chapter-4/2-similarity-search-level-3.ipynb Accuracy calculation on 2048 dim is too slow

1

It is taking ~40 minutes on my CPU which is unbearable. Why not use numpy? `class_ids` was pickled but not used. Here's the full code, takes < 1 second (i...

gitgithan

2 occurences of `feature_list` should be `dataset` instead, also 2048 should be 100. These are defined earlier in ``` num_items = 100000 num_dimensions = 100 dataset = np.random.randn(num_items, num_dimensions) dataset...

gitgithan

``` u = AnnoyIndex(2048) %timeit u.get_nns_by_vector(feature_list[random_image_index], 5, include_distances=True) ``` The above code appeared a few times. `u.load('data/caltech101index.ann')` is missing. Without it all returned indexes are empty. Performance measures are for...

gitgithan

`features = extract_features('cat.jpg', model)` is assuming it is run in colab. I suggest `features = extract_features(IMG_PATH, model)` so it also works locally off colab and uses IMG_PATH which was defined...

gitgithan