train.py

from models import MVDNet as MVDNet
from models import ConsModule as DepthCons
from models import ConsLoss

import argparse
import time
import csv

import numpy as np
import torch
from torch.autograd import Variable
import torch.backends.cudnn as cudnn
import torch.optim
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.data
import custom_transforms
from utils import tensor2array, save_checkpoint, save_path_formatter, adjust_learning_rate
from loss_functions import compute_errors_train, compute_errors_test, compute_angles

from logger import TermLogger, AverageMeter
from itertools import chain
from tensorboardX import SummaryWriter
from data_loader import SequenceFolder

import matplotlib.pyplot as plt
from scipy.misc import imsave
from path import Path
import os
import copy

parser = argparse.ArgumentParser(description='Structure from Motion Learner training on KITTI and CityScapes Dataset',
                                 formatter_class=argparse.ArgumentDefaultsHelpFormatter)

parser.add_argument('data', metavar='DIR',
                    help='path to dataset')
parser.add_argument('--dataset', dest='dataset', default='dataset', metavar='PATH',
                    help='dataset')
parser.add_argument('-j', '--workers', default=4, type=int, metavar='N',
                    help='number of data loading workers')
parser.add_argument('--epochs', default=20, type=int, metavar='N',
                    help='number of total epochs to run')
parser.add_argument('--epoch-size', default=0, type=int, metavar='N',
                    help='manual epoch size (will match dataset size if not set)')
parser.add_argument('-b', '--batch-size', default=16, type=int,
                    metavar='N', help='mini-batch size')
parser.add_argument('--lr', '--learning-rate', default=2e-4, type=float,
                    metavar='LR', help='initial learning rate')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
                    help='momentum for sgd, alpha parameter for adam')
parser.add_argument('--beta', default=0.999, type=float, metavar='M',
                    help='beta parameters for adam')
parser.add_argument('--weight-decay', '--wd', default=0, type=float,
                    metavar='W', help='weight decay')
parser.add_argument('--print-freq', default=10, type=int,
                    metavar='N', help='print frequency')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true',
                    help='evaluate model on validation set')
parser.add_argument('--pretrained-mvdn', dest='pretrained_mvdn', default=None, metavar='PATH',
                    help='path to pre-trained mvdnet model')
parser.add_argument('--pretrained-cons', dest='pretrained_cons', default=None, metavar='PATH',
                    help='path to pre-trained cons_net model')
parser.add_argument('--seed', default=0, type=int, help='seed for random functions, and network initialization')
parser.add_argument('--log-summary', default='progress_log_summary.csv', metavar='PATH',
                    help='csv where to save per-epoch train and valid stats')
parser.add_argument('--log-full', default='progress_log_full.csv', metavar='PATH',
                    help='csv where to save per-gradient descent train stats')
parser.add_argument('--log-output', action='store_true', help='will log dispnet outputs and warped imgs at validation step')
parser.add_argument('--output-print', action='store_true', help='print output depth')
parser.add_argument('--output-dir', default='test_result', type=str, help='Output directory for saving predictions in a numpy files')
parser.add_argument('--ttype', default='train.txt', type=str, help='Text file indicates input data')
parser.add_argument('--ttype2', default='val.txt', type=str, help='Text file indicates input data')
parser.add_argument('--nlabel', type=int ,default=64, help='number of label')
parser.add_argument('--mindepth', type=float ,default=0.5, help='minimum depth')

parser.add_argument('--exp', default='default', type=str, help='Experiment name')
parser.add_argument('-sv', dest='skip_v', action='store_true',
                    help='Skip validation')

parser.add_argument('-nw','--n-weight', type=float ,default=3, help='weight of nmap loss')
parser.add_argument('-dw','--d-weight', type=float ,default=1, help='weight of depth loss')
parser.add_argument('-cw','--c-weight', type=float ,default=0.1, help='weight of cons loss')

parser.add_argument('-tc', dest='train_cons', action='store_true',
                    help='Train Unet for consistency loss')
parser.add_argument('-np', dest='no_pool', action='store_true',
                    help='Less pooling layers in Nnet. Even with more pooling layers, the performance change is insignificant. One of our pretrained models happens to have less pooling layers')

n_iter = 0

def main():
    global n_iter
    args = parser.parse_args()
    save_path = save_path_formatter(args, parser)
    args.save_path = 'checkpoints'/(args.exp+'_'+save_path)
    print('=> will save everything to {}'.format(args.save_path))
    args.save_path.makedirs_p()
    torch.manual_seed(args.seed)

    training_writer = SummaryWriter(args.save_path)
    output_writers = []
    for i in range(3):
        output_writers.append(SummaryWriter(args.save_path/'valid'/str(i)))

    # Data loading code
    normalize = custom_transforms.Normalize(mean=[0.5, 0.5, 0.5],
                                            std=[0.5, 0.5, 0.5])
    
    train_transform = custom_transforms.Compose([
        custom_transforms.RandomScaleCrop(),
        custom_transforms.ArrayToTensor(),
        normalize
    ])


    valid_transform = custom_transforms.Compose([custom_transforms.ArrayToTensor(), normalize])

    
    print("=> fetching scenes in '{}'".format(args.data))
    train_set = SequenceFolder(
        args.data,
        transform=train_transform,
        seed=args.seed,
        ttype=args.ttype,
        dataset = args.dataset
    )
    val_set = SequenceFolder(
        args.data,
        transform=valid_transform,
        seed=args.seed,
        ttype=args.ttype2,
        dataset = args.dataset
    )

    train_set.samples = train_set.samples[:len(train_set) - len(train_set)%args.batch_size]


    print('{} samples found in {} train scenes'.format(len(train_set), len(train_set.scenes)))
    print('{} samples found in {} valid scenes'.format(len(val_set), len(val_set.scenes)))
    train_loader = torch.utils.data.DataLoader(
        train_set, batch_size=args.batch_size, shuffle=True,
        num_workers=args.workers, pin_memory=True)
    val_loader = torch.utils.data.DataLoader(
        val_set, batch_size=1, shuffle=False,
        num_workers=args.workers, pin_memory=True)

    if args.epoch_size == 0:
        args.epoch_size = len(train_loader)

    
    # create model
    print("=> creating model")

    mvdnet = MVDNet(args.nlabel, args.mindepth, no_pool = args.no_pool).cuda()
    mvdnet.init_weights()
    if args.pretrained_mvdn:
        print("=> using pre-trained weights for MVDNet")
        weights = torch.load(args.pretrained_mvdn)   
        mvdnet.load_state_dict(weights['state_dict'])
    
    depth_cons = DepthCons().cuda()
    depth_cons.init_weights()

    if args.pretrained_cons:
        print("=> using pre-trained weights for ConsNet")
        weights = torch.load(args.pretrained_cons)
        depth_cons.load_state_dict(weights['state_dict'])
    
    cons_loss_ = ConsLoss().cuda()
    print('=> setting adam solver')

    if args.train_cons:
        optimizer = torch.optim.Adam(depth_cons.parameters(), args.lr,
                                     betas=(args.momentum, args.beta),
                                     weight_decay=args.weight_decay)
        mvdnet.eval()
    else:
        optimizer = torch.optim.Adam(mvdnet.parameters(), args.lr,
                                     betas=(args.momentum, args.beta),
                                     weight_decay=args.weight_decay)

    cudnn.benchmark = True
    mvdnet = torch.nn.DataParallel(mvdnet)
    depth_cons = torch.nn.DataParallel(depth_cons)

    print(' ==> setting log files')
    with open(args.save_path/args.log_summary, 'w') as csvfile:
        writer = csv.writer(csvfile, delimiter='\t')
        writer.writerow(['train_loss', 'validation_abs_rel', 'validation_abs_diff','validation_sq_rel', 'validation_a1', 'validation_a2', 'validation_a3', 'mean_angle_error'])

    with open(args.save_path/args.log_full, 'w') as csvfile:
        writer = csv.writer(csvfile, delimiter='\t')
        writer.writerow(['train_loss'])


    
    print(' ==> main Loop')
    for epoch in range(args.epochs):
        adjust_learning_rate(args, optimizer, epoch)

        # train for one epoch
        if args.evaluate:
            train_loss = 0
        else:
            train_loss = train(args, train_loader, mvdnet, depth_cons, cons_loss_, optimizer, args.epoch_size, training_writer, epoch)
        if not args.evaluate and (args.skip_v):
            error_names = ['abs_rel', 'abs_diff', 'sq_rel', 'a1', 'a2', 'a3', 'angle']
            errors = [0]*7
        else:
            errors, error_names = validate_with_gt(args, val_loader, mvdnet, depth_cons, epoch, output_writers)

        error_string = ', '.join('{} : {:.3f}'.format(name, error) for name, error in zip(error_names, errors))

        for error, name in zip(errors, error_names):
            training_writer.add_scalar(name, error, epoch)

        # Up to you to chose the most relevant error to measure your model's performance, careful some measures are to maximize (such as a1,a2,a3)
        decisive_error = errors[0]
        with open(args.save_path/args.log_summary, 'a') as csvfile:
            writer = csv.writer(csvfile, delimiter='\t')
            writer.writerow([train_loss, decisive_error, errors[1], errors[2], errors[3], errors[4], errors[5], errors[6]])
        if args.evaluate:
            break
        if args.train_cons:
            save_checkpoint(
                args.save_path, {
                    'epoch': epoch + 1,
                    'state_dict': depth_cons.module.state_dict()
                },
                epoch, file_prefixes = ['cons'])
        else:
            save_checkpoint(
            args.save_path, {
                'epoch': epoch + 1,
                'state_dict': mvdnet.module.state_dict()
            },
            epoch, file_prefixes = ['mvdnet'])


def train(args, train_loader, mvdnet, depth_cons, cons_loss_, optimizer, epoch_size, train_writer, epoch):
    global n_iter
    batch_time = AverageMeter()
    data_time = AverageMeter()
    losses = AverageMeter(precision=4)
    d_losses = AverageMeter(precision=4)
    nmap_losses = AverageMeter(precision=4)
    cons_losses = AverageMeter(precision=4)
    
    # switch to training mode
    if args.train_cons:
        depth_cons.train()
    else:
        mvdnet.train()

    print("Training")
    end = time.time()

    for i, (tgt_img, ref_imgs, gt_nmap, ref_poses, intrinsics, intrinsics_inv, tgt_depth) in enumerate(train_loader):
        # measure data loading time
        data_time.update(time.time() - end)
        tgt_img_var = Variable(tgt_img.cuda())
        ref_imgs_var = [Variable(img.cuda()) for img in ref_imgs]
        gt_nmap_var = Variable(gt_nmap.cuda())
        ref_poses_var = [Variable(pose.cuda()) for pose in ref_poses]
        intrinsics_var = Variable(intrinsics.cuda())
        intrinsics_inv_var = Variable(intrinsics_inv.cuda())
        tgt_depth_var = Variable(tgt_depth.cuda()).cuda()

        # compute output
        pose = torch.cat(ref_poses_var,1)

        
        # get mask
        mask = (tgt_depth_var <= args.nlabel*args.mindepth) & (tgt_depth_var >= args.mindepth) & (tgt_depth_var == tgt_depth_var)
        mask.detach_()
        if mask.any() == 0:
            continue
        
        if args.train_cons:
            with torch.no_grad():
                outputs = mvdnet(tgt_img_var, ref_imgs_var, pose, intrinsics_var, intrinsics_inv_var)
                output_depth1 = outputs[0]
                nmap1 = outputs[1]
        else:
            outputs = mvdnet(tgt_img_var, ref_imgs_var, pose, intrinsics_var, intrinsics_inv_var)
            output_depth1 = outputs[1]
            nmap1 = outputs[2]

        if args.train_cons:
            outputs = depth_cons(output_depth1, nmap1)
            nmap = outputs[:,1:]
            depths = [outputs[:,0]]
        else:
            nmap = nmap1.permute(0,3,1,2) 
            depths = [output_depth1.squeeze(1)]

        loss = 0.
        d_loss = 0.
        nmap_loss = 0.
        cons_loss = 0.
        
        for l, depth in enumerate(depths):
            output = torch.squeeze(depth,1)
            d_loss = d_loss + F.smooth_l1_loss(output[mask], tgt_depth_var[mask])
        
        n_mask = mask.unsqueeze(1).expand(-1,3,-1,-1)
        nmap_loss = nmap_loss + F.smooth_l1_loss(nmap[n_mask], gt_nmap_var[n_mask])

        if args.train_cons:
        	cons_loss =  cons_loss + cons_loss_(depths[-1].unsqueeze(1), tgt_depth_var.unsqueeze(1), nmap.clone(), intrinsics_var, mask.unsqueeze(1))
        	cons_losses.update(cons_loss.item(), args.batch_size)
        loss = loss + args.d_weight*d_loss + args.n_weight*nmap_loss + args.c_weight*cons_loss

        if i > 0 and n_iter % args.print_freq == 0:
            train_writer.add_scalar('total_loss', loss.item(), n_iter)
        # record loss and EPE
        losses.update(loss.item(), args.batch_size)
        d_losses.update(d_loss.item(), args.batch_size)
        nmap_losses.update(nmap_loss.item(), args.batch_size)
        
        
        # compute gradient and do Adam step
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        # measure elapsed time
        batch_time.update(time.time() - end)
        end = time.time()

        with open(args.save_path/args.log_full, 'a') as csvfile:
            writer = csv.writer(csvfile, delimiter='\t')
            writer.writerow([loss.item()])
        if i % args.print_freq == 0:
            print('Train: Time {} Data {} Loss {} NmapLoss {} DLoss {} ConsLoss {}Iter {}/{} Epoch {}/{}'.format(batch_time, data_time, losses, nmap_losses, d_losses, cons_losses, i, len(train_loader), epoch, args.epochs))
        
        if i >= epoch_size - 1:
            break

        n_iter += 1

    return losses.avg[0]


def validate_with_gt(args, val_loader, mvdnet, depth_cons, epoch, output_writers=[]):
    batch_time = AverageMeter()
    error_names = ['abs_rel', 'abs_diff', 'sq_rel', 'a1', 'a2', 'a3', 'mean_angle']
    test_error_names = ['abs_rel','abs_diff','sq_rel','rms','log_rms','a1','a2','a3', 'mean_angle']
    test_error_names1 = ['abs_rel','abs_diff','sq_rel','rms','log_rms','a1','a2','a3', 'mean_angle']
    errors = AverageMeter(i=len(error_names))
    test_errors = AverageMeter(i=len(test_error_names))
    test_errors1 = AverageMeter(i=len(test_error_names1))
    log_outputs = len(output_writers) > 0

    # switch to evaluate mode
    if args.train_cons:
        depth_cons.eval()
    else:
        mvdnet.eval()

    end = time.time()
    with torch.no_grad():
        for i, (tgt_img, ref_imgs, gt_nmap, ref_poses, intrinsics, intrinsics_inv, tgt_depth) in enumerate(val_loader):
            tgt_img_var = Variable(tgt_img.cuda())
            ref_imgs_var = [Variable(img.cuda()) for img in ref_imgs]
            gt_nmap_var = Variable(gt_nmap.cuda())
            ref_poses_var = [Variable(pose.cuda()) for pose in ref_poses]
            intrinsics_var = Variable(intrinsics.cuda())
            intrinsics_inv_var = Variable(intrinsics_inv.cuda())
            tgt_depth_var = Variable(tgt_depth.cuda())

            pose = torch.cat(ref_poses_var,1)
            if (pose != pose).any():
                continue
        
            outputs = mvdnet(tgt_img_var, ref_imgs_var, pose, intrinsics_var, intrinsics_inv_var)
            output_depth = outputs[0]
            output_depth1 = output_depth.clone()
            nmap = outputs[1] 
            nmap1 = nmap.clone()

            output_depth1 = output_depth.clone()
            if args.train_cons:
                outputs = depth_cons(output_depth, nmap.permute(0,3,1,2))
                nmap = outputs[:,1:].permute(0,2,3,1)
                output_depth = outputs[:,0].unsqueeze(1)
            
            mask = (tgt_depth <= args.nlabel*args.mindepth) & (tgt_depth >= args.mindepth) & (tgt_depth == tgt_depth)
            #mask = (tgt_depth <= 10) & (tgt_depth >= args.mindepth) & (tgt_depth == tgt_depth) #for DeMoN testing, to compare against DPSNet you might need to turn on this for fair comparison

            if not mask.any():
                continue

            output_depth1_ = torch.squeeze(output_depth1.data.cpu(),1)
            output_depth_ = torch.squeeze(output_depth.data.cpu(),1)

            
            errors_ = compute_errors_train(tgt_depth, output_depth_, mask)
            test_errors_ = list(compute_errors_test(tgt_depth[mask], output_depth_[mask]))
            test_errors1_ = list(compute_errors_test(tgt_depth[mask], output_depth1_[mask]))

            n_mask = (gt_nmap_var.permute(0,2,3,1)[0,:,:] != 0)
            n_mask = n_mask[:,:,0] | n_mask[:,:,1] | n_mask[:,:,2]
            total_angles_m = compute_angles(gt_nmap_var.permute(0,2,3,1)[0], nmap[0])
            total_angles_m1 = compute_angles(gt_nmap_var.permute(0,2,3,1)[0], nmap1[0])
            
            mask_angles = total_angles_m[n_mask]
            mask_angles1 = total_angles_m1[n_mask]
            total_angles_m[~ n_mask] = 0
            total_angles_m1[~ n_mask] = 0
            errors_.append(torch.mean(mask_angles).item())#/mask_angles.size(0)#[torch.sum(mask_angles).item(), (mask_angles.size(0)),  torch.sum(mask_angles < 7.5).item(), torch.sum(mask_angles < 15).item(), torch.sum(mask_angles < 30).item(), torch.sum(mask_angles < 45).item()]
            test_errors_.append(torch.mean(mask_angles).item())
            test_errors1_.append(torch.mean(mask_angles1).item())
            errors.update(errors_)
            test_errors.update(test_errors_)
            test_errors1.update(test_errors1_)

            # measure elapsed time
            batch_time.update(time.time() - end)
            end = time.time()
            if i % args.print_freq == 0 or i == len(val_loader)-1:
                if args.train_cons:
                    print('valid: Time {} Prev Error {:.4f}({:.4f}) Curr Error {:.4f} ({:.4f}) Prev angle Error {:.4f} ({:.4f}) Curr angle Error {:.4f} ({:.4f}) Iter {}/{}'.format(batch_time, test_errors1.val[0], test_errors1.avg[0], test_errors.val[0], test_errors.avg[0], test_errors1.val[-1], test_errors1.avg[-1], test_errors.val[-1], test_errors.avg[-1], i, len(val_loader)))
                else:
                    print('valid: Time {} Rel Error {:.4f} ({:.4f}) Angle Error {:.4f} ({:.4f}) Iter {}/{}'.format(batch_time, test_errors.val[0], test_errors.avg[0], test_errors.val[-1], test_errors.avg[-1], i, len(val_loader)))
            if args.output_print:
                output_dir= Path(args.output_dir)
                if not os.path.isdir(output_dir):
                    os.mkdir(output_dir)
                plt.imsave(output_dir/'{:04d}_map{}'.format(i,'_dps.png'), output_depth_.numpy()[0], cmap='rainbow')
                np.save(output_dir/'{:04d}{}'.format(i,'_dps.npy'), output_depth_.numpy()[0])
                if args.train_cons:
                	plt.imsave(output_dir/'{:04d}_map{}'.format(i,'_prev.png'), output_depth1_.numpy()[0], cmap='rainbow')
                	np.save(output_dir/'{:04d}{}'.format(i,'_prev.npy'), output_depth1_.numpy()[0])
                # np.save(output_dir/'{:04d}{}'.format(i,'_gt.npy'),tgt_depth.numpy()[0])
                # imsave(output_dir/'{:04d}_aimage{}'.format(i,'.png'), np.transpose(tgt_img.numpy()[0],(1,2,0)))
                # np.save(output_dir/'{:04d}_cam{}'.format(i,'.npy'),intrinsics_var.cpu().numpy()[0])
    if args.output_print:
        np.savetxt(output_dir/args.ttype+'errors.csv', test_errors.avg, fmt='%1.4f', delimiter=',')
        np.savetxt(output_dir/args.ttype+'prev_errors.csv', test_errors1.avg, fmt='%1.4f', delimiter=',')
    return errors.avg, error_names


if __name__ == '__main__':
    main()