crosstyan/OpenGait
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Solve the problem of dimension misuse. (#59)

Browse Source 
* commit for fix dimension

* fix dimension for all method

* restore config

* clean up baseline config

* add contiguous

* rm comment
This commit is contained in:
Junhao Liang
2022-06-28 12:27:16 +08:00
committed by GitHub
parent 715e7448fa
commit 14fa5212d4
14 changed files with 99 additions and 121 deletions
Download Patch File Download Diff File Expand all files Collapse all files
opengait/modeling/models/baseline.py
+6 -11
View File
@@ -19,26 +19,21 @@ class Baseline(BaseModel):
sils = ipts[0]
if len(sils.size()) == 4:
sils = sils.unsqueeze(2)
sils = sils.unsqueeze(1)
del ipts
outs = self.Backbone(sils) # [n, s, c, h, w]
outs = self.Backbone(sils) # [n, c, s, h, w]
# Temporal Pooling, TP
outs = self.TP(outs, seqL, dim=1)[0] # [n, c, h, w]
outs = self.TP(outs, seqL, options={"dim": 2})[0] # [n, c, h, w]
# Horizontal Pooling Matching, HPM
feat = self.HPP(outs) # [n, c, p]
feat = feat.permute(2, 0, 1).contiguous() # [p, n, c]
embed_1 = self.FCs(feat) # [p, n, c]
embed_2, logits = self.BNNecks(embed_1) # [p, n, c]
embed_1 = embed_1.permute(1, 0, 2).contiguous() # [n, p, c]
embed_2 = embed_2.permute(1, 0, 2).contiguous() # [n, p, c]
logits = logits.permute(1, 0, 2).contiguous() # [n, p, c]
embed_1 = self.FCs(feat) # [n, c, p]
embed_2, logits = self.BNNecks(embed_1) # [n, c, p]
embed = embed_1
n, s, _, h, w = sils.size()
n, _, s, h, w = sils.size()
retval = {
'training_feat': {
'triplet': {'embeddings': embed_1, 'labels': labs},
opengait/modeling/models/gaitgl.py
+12 -17
View File
@@ -75,7 +75,7 @@ class GaitGL(BaseModel):
class_num = model_cfg['class_num']
dataset_name = self.cfgs['data_cfg']['dataset_name']
if dataset_name in ['OUMVLP','GREW']:
if dataset_name in ['OUMVLP', 'GREW']:
# For OUMVLP and GREW
self.conv3d = nn.Sequential(
BasicConv3d(1, in_c[0], kernel_size=(3, 3, 3),
@@ -135,12 +135,11 @@ class GaitGL(BaseModel):
self.GLConvB2 = GLConv(in_c[2], in_c[2], halving=3, fm_sign=True, kernel_size=(
3, 3, 3), stride=(1, 1, 1), padding=(1, 1, 1))
self.TP = PackSequenceWrapper(torch.max)
self.HPP = GeMHPP()
self.Head0 = SeparateFCs(64, in_c[-1], in_c[-1])
if 'SeparateBNNecks' in model_cfg.keys():
self.BNNecks = SeparateBNNecks(**model_cfg['SeparateBNNecks'])
self.Bn_head = False
@@ -171,22 +170,18 @@ class GaitGL(BaseModel):
outs = self.GLConvA1(outs)
outs = self.GLConvB2(outs) # [n, c, s, h, w]
outs = self.TP(outs, dim=2, seq_dim=2, seqL=seqL)[0] # [n, c, h, w]
outs = self.TP(outs, seqL=seqL, options={"dim": 2})[0] # [n, c, h, w]
outs = self.HPP(outs) # [n, c, p]
outs = outs.permute(2, 0, 1).contiguous() # [p, n, c]
gait = self.Head0(outs) # [p, n, c]
if self.Bn_head: # Original GaitGL Head
gait = gait.permute(1, 2, 0).contiguous() # [n, c, p]
gait = self.Head0(outs) # [n, c, p]
if self.Bn_head: # Original GaitGL Head
bnft = self.Bn(gait) # [n, c, p]
logi = self.Head1(bnft.permute(2, 0, 1).contiguous()) # [p, n, c]
embed = bnft.permute(0, 2, 1).contiguous() # [n, p, c]
else: # BNNechk as Head
bnft, logi = self.BNNecks(gait) # [p, n, c]
embed = gait.permute(1, 0, 2).contiguous() # [n, p, c]
logi = logi.permute(1, 0, 2).contiguous() # [n, p, c]
logi = self.Head1(bnft) # [n, c, p]
embed = bnft
else: # BNNechk as Head
bnft, logi = self.BNNecks(gait) # [n, c, p]
embed = gait
n, _, s, h, w = sils.size()
retval = {
opengait/modeling/models/gaitpart.py
+12 -13
View File
@@ -45,12 +45,12 @@ class TemporalFeatureAggregator(nn.Module):
def forward(self, x):
"""
Input: x, [n, s, c, p]
Output: ret, [n, p, c]
Input: x, [n, c, s, p]
Output: ret, [n, c, p]
"""
n, s, c, p = x.size()
x = x.permute(3, 0, 2, 1).contiguous() # [p, n, c, s]
feature = x.split(1, 0) # [[n, c, s], ...]
n, c, s, p = x.size()
x = x.permute(3, 0, 1, 2).contiguous() # [p, n, c, s]
feature = x.split(1, 0) # [[1, n, c, s], ...]
x = x.view(-1, c, s)
# MTB1: ConvNet1d & Sigmoid
@@ -73,7 +73,7 @@ class TemporalFeatureAggregator(nn.Module):
# Temporal Pooling
ret = self.TP(feature3x1 + feature3x3, dim=-1)[0] # [p, n, c]
ret = ret.permute(1, 0, 2).contiguous() # [n, p, c]
ret = ret.permute(1, 2, 0).contiguous() # [n, p, c]
return ret
@@ -102,17 +102,16 @@ class GaitPart(BaseModel):
sils = ipts[0]
if len(sils.size()) == 4:
sils = sils.unsqueeze(2)
sils = sils.unsqueeze(1)
del ipts
out = self.Backbone(sils) # [n, s, c, h, w]
out = self.HPP(out) # [n, s, c, p]
out = self.TFA(out, seqL) # [n, p, c]
out = self.Backbone(sils) # [n, c, s, h, w]
out = self.HPP(out) # [n, c, s, p]
out = self.TFA(out, seqL) # [n, c, p]
embs = self.Head(out.permute(1, 0, 2).contiguous()) # [p, n, c]
embs = embs.permute(1, 0, 2).contiguous() # [n, p, c]
embs = self.Head(out) # [n, c, p]
n, s, _, h, w = sils.size()
n, _, s, h, w = sils.size()
retval = {
'training_feat': {
'triplet': {'embeddings': embs, 'labels': labs}
opengait/modeling/models/gaitset.py
+5 -7
View File
@@ -49,30 +49,28 @@ class GaitSet(BaseModel):
ipts, labs, _, _, seqL = inputs
sils = ipts[0] # [n, s, h, w]
if len(sils.size()) == 4:
sils = sils.unsqueeze(2)
sils = sils.unsqueeze(1)
del ipts
outs = self.set_block1(sils)
gl = self.set_pooling(outs, seqL, dim=1)[0]
gl = self.set_pooling(outs, seqL, options={"dim": 2})[0]
gl = self.gl_block2(gl)
outs = self.set_block2(outs)
gl = gl + self.set_pooling(outs, seqL, dim=1)[0]
gl = gl + self.set_pooling(outs, seqL, options={"dim": 2})[0]
gl = self.gl_block3(gl)
outs = self.set_block3(outs)
outs = self.set_pooling(outs, seqL, dim=1)[0]
outs = self.set_pooling(outs, seqL, options={"dim": 2})[0]
gl = gl + outs
# Horizontal Pooling Matching, HPM
feature1 = self.HPP(outs) # [n, c, p]
feature2 = self.HPP(gl) # [n, c, p]
feature = torch.cat([feature1, feature2], -1) # [n, c, p]
feature = feature.permute(2, 0, 1).contiguous() # [p, n, c]
embs = self.Head(feature)
embs = embs.permute(1, 0, 2).contiguous() # [n, p, c]
n, s, _, h, w = sils.size()
n, _, s, h, w = sils.size()
retval = {
'training_feat': {
'triplet': {'embeddings': embs, 'labels': labs}
opengait/modeling/models/gln.py
+7 -9
View File
@@ -89,12 +89,12 @@ class GLN(BaseModel):
sils = ipts[0] # [n, s, h, w]
del ipts
if len(sils.size()) == 4:
sils = sils.unsqueeze(2)
n, s, _, h, w = sils.size()
sils = sils.unsqueeze(1)
n, _, s, h, w = sils.size()
### stage 0 sil ###
sil_0_outs = self.sil_stage_0(sils)
stage_0_sil_set = self.set_pooling(sil_0_outs, seqL, dim=1)[0]
stage_0_sil_set = self.set_pooling(sil_0_outs, seqL, options={"dim": 2})[0]
### stage 1 sil ###
sil_1_ipts = self.MaxP_sil(sil_0_outs)
@@ -105,13 +105,13 @@ class GLN(BaseModel):
sil_2_outs = self.sil_stage_2(sil_2_ipts)
### stage 1 set ###
set_1_ipts = self.set_pooling(sil_1_ipts, seqL, dim=1)[0]
stage_1_sil_set = self.set_pooling(sil_1_outs, seqL, dim=1)[0]
set_1_ipts = self.set_pooling(sil_1_ipts, seqL, options={"dim": 2})[0]
stage_1_sil_set = self.set_pooling(sil_1_outs, seqL, options={"dim": 2})[0]
set_1_outs = self.set_stage_1(set_1_ipts) + stage_1_sil_set
### stage 2 set ###
set_2_ipts = self.MaxP_set(set_1_outs)
stage_2_sil_set = self.set_pooling(sil_2_outs, seqL, dim=1)[0]
stage_2_sil_set = self.set_pooling(sil_2_outs, seqL, options={"dim": 2})[0]
set_2_outs = self.set_stage_2(set_2_ipts) + stage_2_sil_set
set1 = torch.cat((stage_0_sil_set, stage_0_sil_set), dim=1)
@@ -133,11 +133,9 @@ class GLN(BaseModel):
set2 = self.HPP(set2)
set3 = self.HPP(set3)
feature = torch.cat([set1, set2, set3], -
1).permute(2, 0, 1).contiguous()
feature = torch.cat([set1, set2, set3], -1)
feature = self.Head(feature)
feature = feature.permute(1, 0, 2).contiguous() # n p c
# compact_bloack
if not self.pretrain: