sherwoac

import os
import json
import numpy as np


class Intrinsics(object):
    def __init__(self, json_file):
        assert os.path.isfile(json_file), f'{json_file=} not found'
        with open(json_file, 'r') as f:
            intrinsics_data = json.load(f)

sherwoac

import kornia as kn
import kornia.geometry
import kornia.geometry.quaternion
def try_random():
    """
    - for two sets of poses A and B
    - find the rotations between them, A->B = d1
    - find the standard deviation of the components of d rotation
    - sample some random quaternions, C
    - scale the the components according the standard deviation and re-normalize

sherwoac

                # hack to change model config from keras 2->3 compliant
                import h5py
                f = h5py.File(savedModelPath, mode="r+")
                model_config_string = f.attrs.get("model_config")
                if model_config_string.find('"groups": 1,') != -1:
                    model_config_string = model_config_string.replace('"groups": 1,', '')
                    f.attrs.modify('model_config', model_config_string)
                    f.flush()
                    model_config_string = f.attrs.get("model_config")
                    assert model_config_string.find('"groups": 1,') == -1

sherwoac

name: pdf_extractor_env
channels:
  - conda-forge
  - defaults
dependencies:
  - _libgcc_mutex=0.1=main
  - _openmp_mutex=5.1=1_gnu
  - blas=1.0=openblas
  - bottleneck=1.3.5=py39h7deecbd_0
  - brotli=1.0.9=he6710b0_2

sherwoac

name: instant_ngp_env
channels:
  - nvidia
  - conda-forge
  - defaults
dependencies:
  - _libgcc_mutex=0.1=main
  - _openmp_mutex=4.5=1_gnu
  - alsa-lib=1.2.3=h516909a_0
  - blas=1.0=mkl

sherwoac

sherwoac

    def _get_unit_square_intercepts(self, slopes, intercept):
        """
        returns unit square intercepts for given slope (a) and intercepts (b)
        y = ax + b
        solves:
        right: y = a + b
        x = 1
        y = slopes + intercept
        left: y = b
        x = 0

sherwoac

# cf from https://arxiv.org/abs/1812.07035
import tensorflow as tf

def rotation_matrix_to_label_6d_flat(rotation_matrices):
    """
    takes batch x 3x3 rotation matrix, returns a flat batch x 6 floats for loss
    :param rotation_matrices: batch x (3x3) rotation matrix
    :return: flat batch x 6 floats for loss
    """
    return tf.concat([rotation_matrices[:, :, 0], rotation_matrices[:, :, 1]], axis=1)

sherwoac

import numpy as np
import tensorflow as tf
from sklearn.manifold import TSNE
import matplotlib.pyplot as plt
from matplotlib import cm


def explore_embedding_size(number_of_categories, embedding_sizes):
    fig = plt.figure(figsize=(15, 9))
    figure_title = f"Embedding size TSNEs for {number_of_categories} categories"