Doron Adler Norod

Convert Tensorflow/PyTorch models to CoreML with AWS Sagemaker

Intro

CoreML Conversion Demo wiki explains how to:

Prepare Image Classification models using Tensorflow and PyTorch
Convert the models to CoreML format
Run CoreML models on iPad simulator

Software installation

Currently, Neo CoreML conversion service supports models saved in the following frameworks / versions:

Steps to create Textured Mesh from Point Cloud using Meshlab

Get your PointCloud into MeshLab

Import the pointcloud file in ".ply" file format in Meshlab. Before importing make sure you do some pre-processing / cleaning on point cloud so as to ease the process of meshing.

Point Cloud Simplification and Normals Computation

Next we need to reduce the number of point samples for smooth meshing.
- So go to Filters -> Point Set -> Point Cloud Simplification. Enter Number of samples circa 5% of original number of points. Make sure Best Sample Heuristic is checked.
After point cloud simplification, make sure to select Simplified point cloud in the Show Layer Dialog on the right hand side. If not visible, it can be opened by navigating to View -> Show Layer Dialog. Now we need to compute normals for point set.
So go to Filters -> Point Set -> Compute normals for point sets . Enter Neighbour num between 10 - 100. Initially try with 10 and

Kinetics Numeric/String labels

These are mapping files that go between class IDs to class names. These are generated from the training CSV files from each dataset by collecting the unique classes, sorting them, and then numbering them from 0 upwards.

	import os
	import pickle
	import requests

	#download cached token data here:
	#https://drive.google.com/file/d/1g_4w_Re5Y0NmcS2Y55WQzESWDeL2dey6/view?usp=sharing
	#and put it into the same folder as this file

	cachedTokenData = {"maxTokenID":-1,"knownTokenIds":{},"data":[]}
	if os.path.exists("cached_token_data.pickle"):

	# Modified StyleGAN2 Projector with CLIP, addl. losses, kmeans, etc.
	# by Peter Baylies, 2021 -- @pbaylies on Twitter

	# Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
	#
	# NVIDIA CORPORATION and its licensors retain all intellectual property
	# and proprietary rights in and to this software, related documentation
	# and any modifications thereto. Any use, reproduction, disclosure or
	# distribution of this software and related documentation without an express
	# license agreement from NVIDIA CORPORATION is strictly prohibited.

	#
	# ALEPH by Advadnoun, https://colab.research.google.com/drive/1Q-TbYvASMPRMXCOQjkxxf72CXYjR_8Vp
	# "This is a notebook that uses DALL-E's decoder and CLIP to generate images from text. I will very likely make this better & easier to use in the future."
	#
	# rearranged to run locally on faster GPU
	#
	# directions:
	# clone https://github.com/openai/DALL-E/ and https://github.com/openai/CLIP
	# copy relevant files into one dir with this script
	# install torch==1.7.1 and other stuff

	# A Best in Class Checklist
	A boiled down checklist adapted from this [post](https://www.swiftjectivec.com/a-best-in-class-app/), created by @jordanmorgan10.

	> To use this, create a Github Issue in your own repo, and simply copy and paste this text.

	## iOS Core Technology
	_Things any iOS app can benefit from_

	- [ ] iCloud Sync
	- [ ] Focus Filter Support

	# git clone https://github.com/NVlabs/stylegan2
	import os
	import numpy as np
	from scipy.interpolate import interp1d
	from scipy.io import wavfile
	import matplotlib.pyplot as plt
	import PIL.Image
	import moviepy.editor

	import dnnlib

	#!/bin/bash

	# git clone https://github.com/shawwn/stylegan2 -b tpu

	source "$HOME/bin/activate-tf1"

	set -x
	#export TPU_NAME=grpc://0.tcp.ngrok.io:15992
	export TPU_NAME=grpc://0.tcp.ngrok.io:17042
	export GCE_BUCKET=gs://sgappa-multi/stylegan-encoder