Posts on Jam Sylph's little universe

Evolution of CNN

Sun, 01 Mar 2020 00:00:00 +0000

📝 Update Log

2024-05-26: Added latest progress in vision models for 2023-2024, included Phase 6 architecture analysis

2023-11-15: Added analysis of current CNN patterns

2023-09-15: Expanded Phase 5 (2020-present) architecture introduction, added ConvNeXt analysis

2020-03-01: Initial article publication

Evolution of CNN

Phase 1: Foundation Era (1998-2011)

LeNet-5 (1998)

Founder: Yann LeCun

Main Architecture:

7-layer structure: 3 convolutional layers, 2 pooling layers, 2 fully connected layers
Uses 5×5 convolution kernels
Uses sigmoid/tanh activation functions

Breakthroughs:

Complete Analysis of YOLOv8 Decoding Process

Mon, 01 Jan 0001 00:00:00 +0000

Complete Analysis of YOLOv8 Decoding Process

This article provides a detailed analysis of prediction decoding and post-processing mechanisms in the YOLOv8 object detection algorithm, including key components such as DFL (Distribution Focal Loss) decoding and Non-Maximum Suppression (NMS).

1. Prediction Decoding Process (decode_predictions)

YOLOv8 adopts an anchor-free design, where the prediction decoding process converts network outputs into standard bounding box format. The entire process can be divided into several key steps:

Deep Dive into YOLOv5 dataloaders.py

Mon, 01 Jan 0001 00:00:00 +0000

In YOLOv5 object detection tasks, when running the train.py code, the train_loader calls the create_dataloader function, which internally creates an instance of the LoadImagesAndLabels class as the dataset. This create_dataloader function ultimately returns a DataLoader and the dataset:

-dataloaders.py

Working Principle of the `getitem` Method in PyTorch Datasets

The __getitem__ is a special method (magic method) in Python, used in YOLOv5’s LoadImagesAndLabels class to access individual samples from the dataset. This method is called when you use a data loader or directly access the dataset through indexing.

Deep Dive into YOLOv5 Object Detection Code

Mon, 01 Jan 0001 00:00:00 +0000

Deep Dive into YOLOv5 Object Detection Code

This article offers an in-depth analysis of YOLOv5’s training process and data augmentation mechanisms, helping to organize and summarize the internal implementation details of the YOLOv5 object detection model.

1. Analysis of train.py File

1.1 Import Section

import argparse
import math
import os
import random
import subprocess
import sys
import time
from copy import deepcopy
from datetime import datetime, timedelta
from pathlib import Path

try:
    import comet_ml  # must be imported before torch (if installed)
except ImportError:
    comet_ml = None

import numpy as np
import torch
import torch.distributed as dist
import torch.nn as nn
import yaml
from torch.optim import lr_scheduler
from tqdm import tqdm

FILE = Path(__file__).resolve()
ROOT = FILE.parents[0]  # YOLOv5 root directory
if str(ROOT) not in sys.path:
    sys.path.append(str(ROOT))  # add ROOT to PATH
ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative

import val as validate  # for end-of-epoch mAP
from models.experimental import attempt_load
from models.yolo import Model
from utils.autoanchor import check_anchors
from utils.autobatch import check_train_batch_size
from utils.callbacks import Callbacks
from utils.dataloaders import create_dataloader
from utils.downloads import attempt_download, is_url
from utils.general import (
    LOGGER,
    TQDM_BAR_FORMAT,
    check_amp,
    check_dataset,
    check_file,
    check_git_info,
    check_git_status,
    check_img_size,
    check_requirements,
    check_suffix,
    check_yaml,
    colorstr,
    get_latest_run,
    increment_path,
    init_seeds,
    intersect_dicts,
    labels_to_class_weights,
    labels_to_image_weights,
    methods,
    one_cycle,
    print_args,
    print_mutation,
    strip_optimizer,
    yaml_save,
)
from utils.loggers import LOGGERS, Loggers
from utils.loggers.comet.comet_utils import check_comet_resume
from utils.loss import ComputeLoss
from utils.metrics import fitness
from utils.plots import plot_evolve
from utils.torch_utils import (
    EarlyStopping,
    ModelEMA,
    de_parallel,
    select_device,
    smart_DDP,
    smart_optimizer,
    smart_resume,
    torch_distributed_zero_first,
)

LOCAL_RANK = int(os.getenv("LOCAL_RANK", -1))
RANK = int(os.getenv("RANK", -1))
WORLD_SIZE = int(os.getenv("WORLD_SIZE", 1))
GIT_INFO = check_git_info()

1.2 Detailed Explanation of the Train() Function

The Train() function is the core function of YOLOv5 training, responsible for managing the entire training process:

Exploration of the Inception Architecture

Mon, 01 Jan 0001 00:00:00 +0000

📝 Update Log

2024-05-18:

Enhanced the “Adaptive Feature Processing” section, refining the three levels of adaptive mechanisms

Added new “Bottleneck Layer Design Pattern” section, with in-depth analysis of the dimension reduction-processing-expansion design philosophy

2023-10-20:

Comprehensive revision of article structure, strengthening the universality of Inception principles

Added discussion on Inception’s long-term impact on modern network design

2023-05-12:

Expanded on Inception’s influence in other networks

Added application of Inception principles in segmentation networks

2022-08-15:

introduction-to-yolo

Mon, 01 Jan 0001 00:00:00 +0000

introduction-to-yolo

YOLO (You Only Look Once) is a popular real-time object detection algorithm known for its efficient performance and high accuracy. Unlike traditional object detection methods, YOLO treats object detection as a regression problem, directly predicting bounding boxes and class probabilities from the complete image.

Basic Principles of YOLO

The core idea of YOLO is to divide the entire image into an S×S grid, with each grid cell responsible for predicting objects contained within it. Specifically, each grid cell predicts:

Understanding Mamba through MambaStock

Mon, 01 Jan 0001 00:00:00 +0000

📝 Update Log

2024-06-16:

Summarized experiment results and future improvement directions

Added comprehensive comparison analysis with other models

Enhanced technical insights and final conclusions

2024-06-15:

Created MambaStock paper implementation documentation

Detailed analysis of Mamba model architecture

Implemented MambaStock design improvements

Understanding Mamba through MambaStock

1. Review of Mamba Model Architecture

According to the paper, the MambaStock structure is an improvement based on the Mamba model, so let’s first examine the main architecture of the Mamba model (mamba.py file).

Posts on Jam Sylph's little universe

Evolution of CNN

📝 Update Log

Evolution of CNN

Phase 1: Foundation Era (1998-2011)

LeNet-5 (1998)

Complete Analysis of YOLOv8 Decoding Process

Complete Analysis of YOLOv8 Decoding Process

Table of Contents

1. Prediction Decoding Process (decode_predictions)

Deep Dive into YOLOv5 dataloaders.py

-dataloaders.py

Working Principle of the __getitem__ Method in PyTorch Datasets

Deep Dive into YOLOv5 Object Detection Code

Deep Dive into YOLOv5 Object Detection Code

1. Analysis of train.py File

1.1 Import Section

1.2 Detailed Explanation of the Train() Function

Exploration of the Inception Architecture

📝 Update Log

introduction-to-yolo

introduction-to-yolo

Basic Principles of YOLO

Understanding Mamba through MambaStock

📝 Update Log

Understanding Mamba through MambaStock

1. Review of Mamba Model Architecture

Working Principle of the `getitem` Method in PyTorch Datasets