ultralytics 8.0.197 save P, R, F1 curves to metrics (#5354)
Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com> Co-authored-by: erminkev1 <83356055+erminkev1@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Andy <39454881+yermandy@users.noreply.github.com>
This commit is contained in:
Glenn Jocher
GitHub
parent
7fd5dcbd86
commit
12e3eef844
33 changed files with 337 additions and 195 deletions
|
|
@ -307,7 +307,7 @@ class Bottleneck(nn.Module):
|
|||
self.add = shortcut and c1 == c2
|
||||
|
||||
def forward(self, x):
|
||||
"""'forward()' applies the YOLOv5 FPN to input data."""
|
||||
"""'forward()' applies the YOLO FPN to input data."""
|
||||
return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
|
||||
|
||||
|
||||
|
|
|
|||
|
|
@ -192,7 +192,7 @@ class RTDETRDecoder(nn.Module):
|
|||
dropout=0.,
|
||||
act=nn.ReLU(),
|
||||
eval_idx=-1,
|
||||
# training args
|
||||
# Training args
|
||||
nd=100, # num denoising
|
||||
label_noise_ratio=0.5,
|
||||
box_noise_scale=1.0,
|
||||
|
|
@ -225,7 +225,7 @@ class RTDETRDecoder(nn.Module):
|
|||
self.num_queries = nq
|
||||
self.num_decoder_layers = ndl
|
||||
|
||||
# backbone feature projection
|
||||
# Backbone feature projection
|
||||
self.input_proj = nn.ModuleList(nn.Sequential(nn.Conv2d(x, hd, 1, bias=False), nn.BatchNorm2d(hd)) for x in ch)
|
||||
# NOTE: simplified version but it's not consistent with .pt weights.
|
||||
# self.input_proj = nn.ModuleList(Conv(x, hd, act=False) for x in ch)
|
||||
|
|
@ -234,24 +234,24 @@ class RTDETRDecoder(nn.Module):
|
|||
decoder_layer = DeformableTransformerDecoderLayer(hd, nh, d_ffn, dropout, act, self.nl, ndp)
|
||||
self.decoder = DeformableTransformerDecoder(hd, decoder_layer, ndl, eval_idx)
|
||||
|
||||
# denoising part
|
||||
# Denoising part
|
||||
self.denoising_class_embed = nn.Embedding(nc, hd)
|
||||
self.num_denoising = nd
|
||||
self.label_noise_ratio = label_noise_ratio
|
||||
self.box_noise_scale = box_noise_scale
|
||||
|
||||
# decoder embedding
|
||||
# Decoder embedding
|
||||
self.learnt_init_query = learnt_init_query
|
||||
if learnt_init_query:
|
||||
self.tgt_embed = nn.Embedding(nq, hd)
|
||||
self.query_pos_head = MLP(4, 2 * hd, hd, num_layers=2)
|
||||
|
||||
# encoder head
|
||||
# Encoder head
|
||||
self.enc_output = nn.Sequential(nn.Linear(hd, hd), nn.LayerNorm(hd))
|
||||
self.enc_score_head = nn.Linear(hd, nc)
|
||||
self.enc_bbox_head = MLP(hd, hd, 4, num_layers=3)
|
||||
|
||||
# decoder head
|
||||
# Decoder head
|
||||
self.dec_score_head = nn.ModuleList([nn.Linear(hd, nc) for _ in range(ndl)])
|
||||
self.dec_bbox_head = nn.ModuleList([MLP(hd, hd, 4, num_layers=3) for _ in range(ndl)])
|
||||
|
||||
|
|
@ -261,10 +261,10 @@ class RTDETRDecoder(nn.Module):
|
|||
"""Runs the forward pass of the module, returning bounding box and classification scores for the input."""
|
||||
from ultralytics.models.utils.ops import get_cdn_group
|
||||
|
||||
# input projection and embedding
|
||||
# Input projection and embedding
|
||||
feats, shapes = self._get_encoder_input(x)
|
||||
|
||||
# prepare denoising training
|
||||
# Prepare denoising training
|
||||
dn_embed, dn_bbox, attn_mask, dn_meta = \
|
||||
get_cdn_group(batch,
|
||||
self.nc,
|
||||
|
|
@ -278,7 +278,7 @@ class RTDETRDecoder(nn.Module):
|
|||
embed, refer_bbox, enc_bboxes, enc_scores = \
|
||||
self._get_decoder_input(feats, shapes, dn_embed, dn_bbox)
|
||||
|
||||
# decoder
|
||||
# Decoder
|
||||
dec_bboxes, dec_scores = self.decoder(embed,
|
||||
refer_bbox,
|
||||
feats,
|
||||
|
|
@ -316,9 +316,9 @@ class RTDETRDecoder(nn.Module):
|
|||
|
||||
def _get_encoder_input(self, x):
|
||||
"""Processes and returns encoder inputs by getting projection features from input and concatenating them."""
|
||||
# get projection features
|
||||
# Get projection features
|
||||
x = [self.input_proj[i](feat) for i, feat in enumerate(x)]
|
||||
# get encoder inputs
|
||||
# Get encoder inputs
|
||||
feats = []
|
||||
shapes = []
|
||||
for feat in x:
|
||||
|
|
@ -335,13 +335,13 @@ class RTDETRDecoder(nn.Module):
|
|||
def _get_decoder_input(self, feats, shapes, dn_embed=None, dn_bbox=None):
|
||||
"""Generates and prepares the input required for the decoder from the provided features and shapes."""
|
||||
bs = len(feats)
|
||||
# prepare input for decoder
|
||||
# Prepare input for decoder
|
||||
anchors, valid_mask = self._generate_anchors(shapes, dtype=feats.dtype, device=feats.device)
|
||||
features = self.enc_output(valid_mask * feats) # bs, h*w, 256
|
||||
|
||||
enc_outputs_scores = self.enc_score_head(features) # (bs, h*w, nc)
|
||||
|
||||
# query selection
|
||||
# Query selection
|
||||
# (bs, num_queries)
|
||||
topk_ind = torch.topk(enc_outputs_scores.max(-1).values, self.num_queries, dim=1).indices.view(-1)
|
||||
# (bs, num_queries)
|
||||
|
|
@ -352,7 +352,7 @@ class RTDETRDecoder(nn.Module):
|
|||
# (bs, num_queries, 4)
|
||||
top_k_anchors = anchors[:, topk_ind].view(bs, self.num_queries, -1)
|
||||
|
||||
# dynamic anchors + static content
|
||||
# Dynamic anchors + static content
|
||||
refer_bbox = self.enc_bbox_head(top_k_features) + top_k_anchors
|
||||
|
||||
enc_bboxes = refer_bbox.sigmoid()
|
||||
|
|
@ -373,7 +373,7 @@ class RTDETRDecoder(nn.Module):
|
|||
# TODO
|
||||
def _reset_parameters(self):
|
||||
"""Initializes or resets the parameters of the model's various components with predefined weights and biases."""
|
||||
# class and bbox head init
|
||||
# Class and bbox head init
|
||||
bias_cls = bias_init_with_prob(0.01) / 80 * self.nc
|
||||
# NOTE: the weight initialization in `linear_init_` would cause NaN when training with custom datasets.
|
||||
# linear_init_(self.enc_score_head)
|
||||
|
|
|
|||
|
|
@ -81,7 +81,7 @@ class AIFI(TransformerEncoderLayer):
|
|||
"""Forward pass for the AIFI transformer layer."""
|
||||
c, h, w = x.shape[1:]
|
||||
pos_embed = self.build_2d_sincos_position_embedding(w, h, c)
|
||||
# flatten [B, C, H, W] to [B, HxW, C]
|
||||
# Flatten [B, C, H, W] to [B, HxW, C]
|
||||
x = super().forward(x.flatten(2).permute(0, 2, 1), pos=pos_embed.to(device=x.device, dtype=x.dtype))
|
||||
return x.permute(0, 2, 1).view([-1, c, h, w]).contiguous()
|
||||
|
||||
|
|
@ -213,7 +213,7 @@ class MSDeformAttn(nn.Module):
|
|||
if d_model % n_heads != 0:
|
||||
raise ValueError(f'd_model must be divisible by n_heads, but got {d_model} and {n_heads}')
|
||||
_d_per_head = d_model // n_heads
|
||||
# you'd better set _d_per_head to a power of 2 which is more efficient in our CUDA implementation
|
||||
# Better to set _d_per_head to a power of 2 which is more efficient in a CUDA implementation
|
||||
assert _d_per_head * n_heads == d_model, '`d_model` must be divisible by `n_heads`'
|
||||
|
||||
self.im2col_step = 64
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue