visualize_features.py

#
#
#      0=================================0
#      |    Kernel Point Convolutions    |
#      0=================================0
#
#
# ----------------------------------------------------------------------------------------------------------------------
#
#      Callable script to test any model on any dataset
#
# ----------------------------------------------------------------------------------------------------------------------
#
#      Hugues THOMAS - 11/06/2018
#


# ----------------------------------------------------------------------------------------------------------------------
#
#           Imports and global variables
#       \**********************************/
#

# Common libs
import time
import os
import numpy as np

# My libs
from utils.config import Config
from utils.visualizer import ModelVisualizer
from models.KPCNN_model import KernelPointCNN
from models.KPFCNN_model import KernelPointFCNN

# Datasets
from datasets.ModelNet40 import ModelNet40Dataset
from datasets.ShapeNetPart import ShapeNetPartDataset
from datasets.S3DIS import S3DISDataset
from datasets.Semantic3D import Semantic3DDataset
from datasets.NPM3D import NPM3DDataset
from datasets.Scannet import ScannetDataset


# ----------------------------------------------------------------------------------------------------------------------
#
#           Utility functions
#       \***********************/
#


def visu_caller(path, step_ind, relu_idx, compute_activations):

    # Check if activation have already been computed
    if relu_idx is not None:
        visu_path = os.path.join('visu',
                                 'visu_' + path.split('/')[-1],
                                 'top_activations',
                                 'Relu{:02d}'.format(relu_idx))
        if not os.path.exists(visu_path):
            message = 'No activations found for Relu number {:d} of the model {:s}.'
            print(message.format(relu_idx, path.split('/')[-1]))
            compute_activations = True
        else:
            # Get the list of files
            feature_files = np.sort([f for f in os.listdir(visu_path) if f.endswith('.ply')])
            if len(feature_files) == 0:
                message = 'No activations found for Relu number {:d} of the model {:s}.'
                print(message.format(relu_idx, path.split('/')[-1]))
                compute_activations = True
    else:
        compute_activations = True

    if compute_activations:

        ##########################
        # Initiate the environment
        ##########################

        # Choose which gpu to use
        GPU_ID = '0'

        # Set GPU visible device
        os.environ['CUDA_VISIBLE_DEVICES'] = GPU_ID

        # Disable warnings
        os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'

        ###########################
        # Load the model parameters
        ###########################

        # Load model parameters
        config = Config()
        config.load(path)

        ##################################
        # Change model parameters for test
        ##################################

        # Change parameters for the test here. For example, you can stop augmenting the input data.

        #config.augment_noise = 0.0001
        #config.augment_symmetries = False

        config.batch_num = 3
        config.in_radius = 4
        config.validation_size = 200

        ##############
        # Prepare Data
        ##############

        print()
        print('Dataset Preparation')
        print('*******************')

        # Initiate dataset configuration
        if config.dataset.startswith('ModelNet40'):
            dataset = ModelNet40Dataset(config.input_threads)
        elif config.dataset == 'S3DIS':
            dataset = S3DISDataset(config.input_threads)
            on_val = True
        elif config.dataset == 'Scannet':
            dataset = ScannetDataset(config.input_threads, load_test=True)
        elif config.dataset.startswith('ShapeNetPart'):
            dataset = ShapeNetPartDataset(config.dataset.split('_')[1], config.input_threads)
        elif config.dataset == 'NPM3D':
            dataset = NPM3DDataset(config.input_threads, load_test=True)
        elif config.dataset == 'Semantic3D':
            dataset = Semantic3DDataset(config.input_threads)
        else:
            raise ValueError('Unsupported dataset : ' + config.dataset)

        # Create subsample clouds of the models
        dl0 = config.first_subsampling_dl
        dataset.load_subsampled_clouds(dl0)

        # Initialize input pipelines
        if config.dataset == 'S3DIS':
            dataset.init_input_pipeline(config)
        else:
            dataset.init_test_input_pipeline(config)

        ##############
        # Define Model
        ##############

        print('Creating Model')
        print('**************\n')
        t1 = time.time()

        if config.dataset.startswith('ShapeNetPart'):
            model = KernelPointFCNN(dataset.flat_inputs, config)
        elif config.dataset.startswith('S3DIS'):
            model = KernelPointFCNN(dataset.flat_inputs, config)
        elif config.dataset.startswith('Scannet'):
            model = KernelPointFCNN(dataset.flat_inputs, config)
        elif config.dataset.startswith('NPM3D'):
            model = KernelPointFCNN(dataset.flat_inputs, config)
        elif config.dataset.startswith('ModelNet40'):
            model = KernelPointCNN(dataset.flat_inputs, config)
        elif config.dataset.startswith('Semantic3D'):
            model = KernelPointFCNN(dataset.flat_inputs, config)
        else:
            raise ValueError('Unsupported dataset : ' + config.dataset)

        # Find all snapshot in the chosen training folder
        snap_path = os.path.join(path, 'snapshots')
        snap_steps = [int(f[:-5].split('-')[-1]) for f in os.listdir(snap_path) if f[-5:] == '.meta']

        # Find which snapshot to restore
        chosen_step = np.sort(snap_steps)[step_ind]
        chosen_snap = os.path.join(path, 'snapshots', 'snap-{:d}'.format(chosen_step))

        # Create a tester class
        visualizer = ModelVisualizer(model, restore_snap=chosen_snap)
        t2 = time.time()

        print('\n----------------')
        print('Done in {:.1f} s'.format(t2 - t1))
        print('----------------\n')

        #####################
        # Start visualization
        #####################

        print('Start visualization')
        print('*******************\n')

        relu_idx = visualizer.top_relu_activations(model, dataset, relu_idx)

    # Show the computed activations
    ModelVisualizer.show_activation(path, relu_idx)

# ----------------------------------------------------------------------------------------------------------------------
#
#           Main Call
#       \***************/
#


if __name__ == '__main__':

    ###############################
    # Choose the model to visualize
    ###############################

    #
    #   Here you can choose which model you want to test with the variable test_model. Here are the possible values :
    #
    #       > 'last_ModelNet40': Automatically retrieve the last trained model on ModelNet40
    #
    #       > 'last_ShapeNetPart': Automatically retrieve the last trained model on ShapeNetPart
    #
    #       > 'last_S3DIS': Automatically retrieve the last trained model on S3DIS
    #
    #       > 'results/Log_YYYY-MM-DD_HH-MM-SS': Directly provide the path of a trained model
    #

    # S3DIS
    chosen_log = 'results/Log_2019-03-19_19-14-24'  # => normal KPConv

    #
    #   You can also choose the index of the snapshot to load (last by default)
    #

    chosen_snapshot = -1

    #
    #   Eventually you can choose which feature is visualized (index of the deform operation in the network).
    #   Let it be None to choose later.
    #

    chosen_relu = 0

    #
    #   Because of the time needed to compute feature activations for the test set, if you already computed them, they
    #   are saved and used again. Set this parameter to True if you want to compute new activations and erase the old
    #   ones. N.B. if chosen_relu = None, the code always recompute activations. Chose a relu idx to avoid it.
    #

    compute_activations = False

    #
    #   If you want to modify certain parameters in the Config class, for example, to stop augmenting the input data,
    #   there is a section for it in the function "test_caller" defined above.
    #

    ###########################
    # Call the test initializer
    ###########################

    # Automatically retrieve the last trained model
    if chosen_log in ['last_ModelNet40', 'last_ShapeNetPart', 'last_S3DIS']:

        # Dataset name
        test_dataset = '_'.join(chosen_log.split('_')[1:])

        # List all training logs
        logs = np.sort([os.path.join('results', f) for f in os.listdir('results') if f.startswith('Log')])

        # Find the last log of asked dataset
        for log in logs[::-1]:
            log_config = Config()
            log_config.load(log)
            if log_config.dataset.startswith(test_dataset):
                chosen_log = log
                break

        if chosen_log in ['last_ModelNet40', 'last_ShapeNetPart', 'last_S3DIS']:
            raise ValueError('No log of the dataset "' + test_dataset + '" found')

    # Check if log exists
    # if not os.path.exists(chosen_log):
    #     raise ValueError('The given log does not exists: ' + chosen_log)

    # Let's go
    visu_caller(chosen_log, chosen_snapshot, chosen_relu, compute_activations)