Some useful links I find here and there

Machine_Learning.md

About Gist https://www.labnol.org/internet/github-gist-tutorial/28499/

(blog)SVMs
http://blog.hackerearth.com/simple-tutorial-svm-parameter-tuning-python-r

Background subtraction
https://github.com/stgstg27/Background-Subtraction

==Visualizaion==

Basic data visualization maps
https://mubaris.com/2017/09/26/introduction-to-data-visualizations-using-python/

More visualisation
https://github.com/ContextLab/storytelling-with-data/blob/master/data-stories/education/tutorial.ipynb

Visualisation with pandas
https://www.kaggle.com/residentmario/univariate-plotting-with-pandas/notebook

Average Precision as AU-PR curve
https://sanchom.wordpress.com/tag/average-precision/

https://medium.com/@jonathan_hui/map-mean-average-precision-for-object-detection-45c121a31173

==TSNE==

Misreading TSNE plots
https://distill.pub/2016/misread-tsne/

==CNN/DL==

SGD >> Adam for Generalisation:
https://arxiv.org/abs/1705.08292

https://shaoanlu.wordpress.com/2017/05/29/sgd-all-which-one-is-the-best-optimizer-dogs-vs-cats-toy-experiment/

CS231n Gradient check http://cs231n.github.io/neural-networks-3/#gradcheck

Open Images dataset maker https://github.com/aferriss/openImageDownloader

DL Mistakes http://ppwwyyxx.com/2017/Unawareness-Of-Deep-Learning-Mistakes/#more

Training classification network- kaggle10th https://towardsdatascience.com/image-classification-challenge-using-transfer-learning-and-deep-learning-studio-2e89c3189fcf

Kaggle4th classification tips https://www.kaggle.com/c/cdiscount-image-classification-challenge/discussion/45733

Kaggle #1 classification tips https://medium.com/neuralspace/kaggle-1-winning-approach-for-image-classification-challenge-9c1188157a86

Warm restarts paper https://arxiv.org/pdf/1608.03983.pdf

Optimal Learning rate https://towardsdatascience.com/estimating-optimal-learning-rate-for-a-deep-neural-network-ce32f2556ce0

Hyperparams https://towardsdatascience.com/artificial-intelligence-hyperparameters-48fa29daa516

On Convolutional NN

(blog)About CNN developments through the years- https://adeshpande3.github.io/adeshpande3.github.io/The-9-Deep-Learning-Papers-You-Need-To-Know-About.html

Understanding Convolutional operation in CNN- https://ujjwalkarn.me/2016/08/11/intuitive-explanation-convnets/

VIsualising MNIST & understanding Dimensionality Reduction http://colah.github.io/posts/2014-10-Visualizing-MNIST/

Lao ML notes http://claoudml.strikingly.com

Tfrecords https://planspace.org/20170323-tfrecords_for_humans/

https://planspace.org/20170403-images_and_tfrecords/

Google ML crash course https://developers.google.com/machine-learning/crash-course/ml-intro

(blog)R-CNN to Mask R-CNN https://blog.athelas.com/a-brief-history-of-cnns-in-image-segmentation-from-r-cnn-to-mask-r-cnn-34ea83205de4

(blog)Fast, Faster R-CNN https://tryolabs.com/blog/2018/01/18/faster-r-cnn-down-the-rabbit-hole-of-modern-object-detection/

https://jhui.github.io/2017/03/15/Fast-R-CNN-and-Faster-R-CNN/

(blog)RoI pooling https://blog.deepsense.ai/region-of-interest-pooling-explained/

(blog)Receptive field arithmetic https://medium.com/mlreview/a-guide-to-receptive-field-arithmetic-for-convolutional-neural-networks-e0f514068807

(SO answer) Anchors and faster-RCNN https://stats.stackexchange.com/questions/265875/anchoring-faster-rcnn

(SO Answer) Cnn filter weights initialization https://stats.stackexchange.com/questions/200513/how-to-initialize-the-elements-of-the-filter-matrix

(blog)cross entropy http://rdipietro.github.io/friendly-intro-to-cross-entropy-loss/

(article)Cross Entropy losses - categorical, focal https://gombru.github.io/2018/05/23/cross_entropy_loss/

(article)Transfer Learning/Fine tune CNN http://cs231n.github.io/transfer-learning/

(coursera)CNN/NMS/Object Detection https://www.coursera.org/learn/convolutional-neural-networks/lecture/dvrjH/non-max-suppression

(medium article)Image augmentation with tf https://medium.com/ymedialabs-innovation/data-augmentation-techniques-in-cnn-using-tensorflow-371ae43d5be9

https://towardsdatascience.com/image-augmentation-for-deep-learning-using-keras-and-histogram-equalization-9329f6ae5085

Also an augmentor library https://github.com/mdbloice/Augmentor

On Bounding Box Regression https://pdxscholar.library.pdx.edu/cgi/viewcontent.cgi?article=4949&context=open_access_etds

Siamese Network Image similarity

Tensorflow series https://blog.metaflow.fr/tensorflow-a-primer-4b3fa0978be3

==PYTHON==

Python Tutorial https://www.python-course.eu/

Guide to import https://chrisyeh96.github.io/2017/08/08/definitive-guide-python-imports.html

Amazing things about python https://nedbatchelder.com/text/names.html https://stackoverflow.com/questions/5131538/slicing-a-list-in-python-without-generating-a-copy

Why self is here to stay http://neopythonic.blogspot.in/2008/10/why-explicit-self-has-to-stay.html

The init self confusion https://stackoverflow.com/questions/625083/python-init-and-self-what-do-they-do

(website)Learn Python http://python.net/~goodger/projects/pycon/2007/idiomatic/handout.html

==KAGGLE==

kaggle ensemble guide https://mlwave.com/kaggle-ensembling-guide/

(blog)Kaggle Zoo Solution http://benanne.github.io/2014/04/05/galaxy-zoo.html

Setting up the computer https://www.kaggle.com/c/allstate-claims-severity/discussion/26423#150025

Interesting thing about fully conv nets like Retinanet - resizing image during preprocessing (default 800x1333) -
check min side, if min size < 800 then upsample to 800 and note the scale, then scale max size with same scale, if this exceeds 1333 then scale to 1333 and note this new scale as final scale- finally resize image both sides with this final scale.
if min size > 800 then downsample to 800 and rest same procedure.

smallest_side = min(rows, cols)

# rescale the image so the smallest side is min_side
scale = min_side / smallest_side

# check if the largest side is now greater than max_side, which can happen
# when images have a large aspect ratio
largest_side = max(rows, cols)
if largest_side * scale > max_side:
    scale = max_side / largest_side

Basically, all images are squashed to this size

On AI companies
https://a16z.com/2020/02/16/the-new-business-of-ai-and-how-its-different-from-traditional-software/