xxq250
/
mindspore
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix bug v0.2

This commit is contained in:
xinjun ma 2021-06-22 20:14:10 +08:00
parent 3bc14d1c3e
commit 169b4b587e
2 changed files with 15 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
model_zoo/research/cv/eppmvsnet/src/blendedmvs.py
View File

@ -14,18 +14,18 @@
# ============================================================================
"""blendedmvs dataset"""
from collections import defaultdict
from object import Compose
from PIL import Image
from copy import deepcopy
from src.utils import read_pfm
import os
import cv2
from collections import defaultdict
from copy import deepcopy
import cv2
import numpy as np
from PIL import Image
import mindspore.dataset.vision.py_transforms as py_vision
from src.utils import read_pfm
class Compose:
"""Composes several transforms together.
@ -59,6 +59,7 @@ class Compose:
class BlendedMVSDataset:
"""blendedmvs dataset"""
def __init__(self, root_dir, split, n_views=3, levels=3, depth_interval=128.0, img_wh=(768, 576),
crop_wh=(640, 512), scale=False, scan=None, training_tag=False):
"""

15
model_zoo/research/cv/eppmvsnet/src/eppmvsnet.py
View File

@ -110,7 +110,7 @@ class SingleStageP1(nn.Cell):
self.D = Tensor(self.depth_number - 1, mstype.float32)
def construct(self, sample, depth_num, depth_start_override=None, depth_interval_override=None, timing=True):
"""construct fuction of part1 of single stage 1"""
"""construct function of part1 of single stage 1"""
ref_feat, _, _ = sample
depth_start = depth_start_override # n111 or n1hw
depth_interval = depth_interval_override # n111
@ -407,8 +407,7 @@ class EPPMVSNetP1(nn.Cell):
depth_num=self.n_depths[0],
depth_start_override=depth_start,
depth_interval_override=depth_interval *
self.interval_ratios[
0])
self.interval_ratios[0])
_, src_feats, proj_mats = [ref_feat_1, srcs_feat_1, proj_mats[:, :, 2]]
pixel_distances = []
@ -451,7 +450,7 @@ class EPPMVSNetP3(nn.Cell):
proj_mats[:, :, 2]],
self.n_depths[0])
stage2_conf_interval = self.shrink_ratio * conf_range_1 / self.n_depths[0] * (
depth_interval * self.interval_ratios[0] * self.n_depths[0]) / self.n_depths[1]
depth_interval * self.interval_ratios[0] * self.n_depths[0]) / self.n_depths[1]
else:
est_depth_1, _, _ = self.stage1_p3(cost_volume_list_stage1,
depth_values_stage1, [ref_feat_1, srcs_feat_1,
@ -474,7 +473,7 @@ class EPPMVSNetP3(nn.Cell):
depth_interval_override=stage2_conf_interval,
uncertainty_maps=uncertainty_maps_2)
stage3_conf_interval = self.shrink_ratio * conf_range_2 / self.n_depths[1] * (
stage2_conf_interval * self.n_depths[1]) / self.n_depths[2]
stage2_conf_interval * self.n_depths[1]) / self.n_depths[2]
else:
est_depth_2, _, _ = self.stage2([ref_feat_2, srcs_feat_2, proj_mats[:, :, 1]],
depth_num=self.n_depths[1], depth_start_override=depth_start_2,
@ -489,17 +488,17 @@ class EPPMVSNetP3(nn.Cell):
depth_start_3 = P.ResizeBilinear((H // 2, W // 2), False)(est_depth_2)
if self.entropy_range:
est_depth_3, prob_map_3, pair_results_3, _ = self.stage3([ref_feat_3, srcs_feat_3, proj_mats[:, :, 0]],
est_depth_3, prob_map_3, _, _ = self.stage3([ref_feat_3, srcs_feat_3, proj_mats[:, :, 0]],
depth_num=self.n_depths[2],
depth_start_override=depth_start_3,
depth_interval_override=stage3_conf_interval,
uncertainty_maps=uncertainty_maps_3)
else:
est_depth_3, prob_map_3, pair_results_3 = self.stage3([ref_feat_3, srcs_feat_3, proj_mats[:, :, 0]],
est_depth_3, prob_map_3, _ = self.stage3([ref_feat_3, srcs_feat_3, proj_mats[:, :, 0]],
depth_num=self.n_depths[2],
depth_start_override=depth_start_3,
depth_interval_override=depth_interval *
self.interval_ratios[2],
self.interval_ratios[2],
uncertainty_maps=uncertainty_maps_3)
refined_depth = est_depth_3
return refined_depth, prob_map_3