wch
/
brep2sdf
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Commit Graph

  • 87a1654d03 暂存 final mckay 2025-05-24 20:16:31 +0800
  • ff2cf97431 loss 分支train mckay 2025-05-23 16:23:52 +0800
  • 73e7dc9fd9 可以零表面 normal 降 mckay 2025-05-23 16:02:06 +0800
  • 60abaf6aeb cuda加速encoder forward，但是效果不好，做保存 mckay 2025-05-23 15:32:31 +0800
  • d266c37a04 stage1 使用完整loss mckay 2025-05-20 22:53:52 +0800
  • c9aadb2d0a 背景 normal 有提升 mckay 2025-05-20 18:47:42 +0800
  • 4c7b91ffdc dot mckay 2025-05-10 13:53:13 +0800
  • f49da7c255 梯度计算 安全归一化 ， 梯度可能会很小 mckay 2025-05-09 19:33:10 +0800
  • 2109c4a6f3 有形状了 mckay 2025-05-08 01:06:59 +0800
  • 6c274cd293 替换stage3 训练数据 mckay 2025-05-07 15:52:44 +0800
  • 1297c3c28e 只用忙的训练stage1 mckay 2025-05-07 15:44:20 +0800
  • 53a2df8522 训练数据打乱 mckay 2025-05-07 15:43:16 +0800
  • 3e2beac722 尝试降损失 mckay 2025-05-07 02:38:13 +0800
  • b5d986ecdd 加了全局采样训练 mckay 2025-05-06 20:37:48 +0800
  • 3eca153112 脚本修改 mckay 2025-05-06 18:04:17 +0800
  • 8ae855f411 stage2 训练改batch mckay 2025-05-06 12:59:45 +0800
  • 965c0864c3 可以三阶段训练，但是 stage2 未使用 batch，会溢出显存 mckay 2025-05-06 12:41:58 +0800
  • 9689c0314e 可以二阶段训练，但是二阶段并行能力很差 mckay 2025-05-05 19:22:58 +0800
  • ca9a56b198 修复min max mckay 2025-05-05 19:21:56 +0800
  • 1f92faad72 优化动态填充，避免warning mckay 2025-05-05 16:05:08 +0800
  • 037f1236e1 train psdf回退 mckay 2025-05-05 14:03:13 +0800
  • 376d078d20 非流形于psdf分开计算 mckay 2025-05-03 20:37:21 +0800
  • 794b2bb220 脚本新增 mckay 2025-05-03 20:27:59 +0800
  • 2bed5b89cd 背景场训练 mckay 2025-05-03 20:27:18 +0800
  • d86989454d 修复patch min max处理 mckay 2025-05-03 20:20:13 +0800
  • 81742489ca fix:之前采样点集中在边缘，现在是表面均匀采样 mckay 2025-04-30 15:15:58 +0800
  • 4ef6d1638f 训练sample修改 mckay 2025-04-30 00:44:39 +0800
  • f7d79eaf52 参数调整 mckay 2025-04-30 00:44:08 +0800
  • 061ddfeda0 优化net计算 mckay 2025-04-28 20:25:41 +0800
  • d7227b91a2 增加伪sdf loss mckay 2025-04-28 19:33:35 +0800
  • 0c40f0f5e3 改了优化器和学习率策略 mckay 2025-04-28 17:56:02 +0800
  • 19dfe8fcc7 增加非流型点训练 mckay 2025-04-28 14:35:24 +0800
  • 1cbf7ffffc 优化encoder farward 循环 mckay 2025-04-28 14:34:26 +0800
  • 6b53d8b1bf 八叉树优化： 1. 去掉 parent_indices索引和child_indices，改用 8*i+1,8*i+8来索引child 2.改变叶节点查找，使用层级遍历方式查找更新叶节点的方式。（为了利用torch的并行计算功能） mckay 2025-04-27 20:36:19 +0800
  • 09986f74fa 增加非流型点loss处理 mckay 2025-04-26 19:10:36 +0800
  • dd838ea22f 流型损失可以训练，法线损失有问题 mckay 2025-04-25 23:55:05 +0800
  • 58e000dad3 可以torch jit script mckay 2025-04-25 21:33:50 +0800
  • f98c9ac227 可以训练，但是和torch jit还有一些兼容问题 mckay 2025-04-25 18:30:55 +0800
  • fa17441396 优化了八叉树并行，确保一个叶节点不超过两个面 mckay 2025-04-23 20:31:23 +0800
  • b89989bbc1 device相关优化，但是显存峰值问题还在 mckay 2025-04-22 17:32:44 +0800
  • 56efa123ec 暂时可用的ISG mckay 2025-04-22 15:23:43 +0800
  • f0af2208a9 octree单独作为一个模块，且解决显存不足问题 mckay 2025-04-21 21:18:20 +0800
  • 595f24effa 显存不够，待优化 mckay 2025-04-21 18:13:41 +0800
  • 0d0f6ecfca ruff 清理包导入 mckay 2025-04-21 13:35:45 +0800
  • 2bb0864190 加入邻接图作为八叉树subdivide依据 mckay 2025-04-21 13:25:07 +0800
  • 8173bae5d5 loss删除无用代码 mckay 2025-04-18 15:22:04 +0800
  • e20686242f 优化八叉树encoder，使用long直接索引 特征向量 mckay 2025-04-18 13:20:04 +0800
  • 18b0ed99c8 拆分数据预处理 mckay 2025-04-18 13:13:22 +0800
  • 8466bd199c 更改ltorch 版本 2.6.0 mckay 2025-04-18 13:12:29 +0800
  • 6a1e871e7b 优化训练脚本 mckay 2025-04-17 15:12:08 +0800
  • 123e21a9af 增加c++ ISOsurfacing mckay 2025-04-17 14:56:01 +0800
  • a3a2a2c7a1 增加非零表面采样点，优化batch train mckay 2025-04-16 21:06:05 +0800
  • 98372f1717 优化八叉树性能 mckay 2025-04-16 15:03:40 +0800
  • 48755817c0 Revert "oct 修改中间版本" mckay 2025-04-17 15:02:42 +0800
  • 5bd4a8d866 oct 修改中间版本 mckay 2025-04-15 14:18:11 +0800
  • 21f8369e20 兼容torch.jit.script，load和save方便。但是牺牲了随机访问，性能降了很多 mckay 2025-04-10 12:47:07 +0800
  • 2ac55ea3df 基本可以跑通，除了八叉树部分保存还有问题 mckay 2025-04-09 14:41:01 +0800
  • 26e3984f76 加强gen name list功能 mckay 2025-04-09 14:38:44 +0800
  • d5ac19e093 调c进行isosurfing mckay 2025-04-08 15:26:55 +0800
  • cc88dfb798 支持normal mckay 2025-04-06 20:53:14 +0800
  • 1e4c360403 可以加载和保存 mckay 2025-04-03 22:37:11 +0800
  • 32d4e45196 修改预处理 mckay 2025-04-03 16:42:31 +0800
  • a7541c9da3 优化写法，兼容 torch mckay 2025-04-03 16:42:12 +0800
  • b33203a7e6 fix: logger mckay 2025-03-31 17:13:25 +0800
  • 8fe07b2b3a gitignore 还有config 修改 mckay 2025-03-31 15:25:22 +0800
  • 845ad864b5 引入 NH-Rep 从 mesh 采样二面角的处理程序 mckay 2025-03-31 15:23:24 +0800
  • 2e8dc3b0fe 可以直接从 step 开始处理的版本 mckay 2025-03-31 15:21:42 +0800
  • 7f2049c4bd 加保存checkpoint mckay 2025-03-27 18:02:11 +0800
  • 584354bbf7 能在八叉树上面训练，但是速度很慢 mckay 2025-03-24 21:50:09 +0800
  • cabc98090e BREAKPOINT: first attremp based on brepgen encoder and deepsef decoder mckay 2025-03-23 15:43:51 +0800
  • 7efcfaae0c feat: Add base tracer and rendering utilities for ray tracing NH-Rep mckay 2025-03-07 19:07:55 +0800
  • 03b2858e81 feat: Add TorchScript model tracing and export functionality mckay 2025-03-07 19:05:22 +0800
  • 2f83f42e5b feat: Enhance training pipeline and configuration management mckay 2025-03-03 21:53:04 +0800
  • c114115239 log change mckay 2025-02-23 19:35:31 +0800
  • d42b3b46dd NOTE it's base mckay 2025-02-23 19:35:00 +0800
  • ee7bf687b2 feat: Add advanced loss computation methods to LossManager mckay 2025-02-22 15:32:32 +0800
  • a874ea78fb feat: Enhance LossManager with advanced loss computation methods mckay 2025-02-19 23:56:31 +0800
  • 6be600a52f feat: Add LossManager for manifold loss computation mckay 2025-02-19 21:54:07 +0800
  • 0ff9f5c5b9 gitignore mckay 2025-02-18 21:46:50 +0800
  • a40fb611d5 feat: Implement NHREP Dataset and Training Pipeline mckay 2025-02-18 21:45:47 +0800
  • 73d48d23e3 refactor: Adjust logger default settings and error handling mckay 2025-02-18 21:45:33 +0800
  • 14674e1a80 refactor: simplest trrain loss mckay 2025-02-16 23:36:35 +0800
  • 285aaf86dd refactor: Add detailed comments and improve code readability in run.py training method mckay 2025-02-16 17:43:43 +0800
  • 39f56470ef refactor: Replace print statements with logger in run.py and network.py mckay 2025-02-16 16:20:16 +0800
  • 77c4e0dc37 feat: Add configurable CSV output filename for evaluation results mckay 2025-02-15 18:15:28 +0800
  • 5a708af310 docs: preprocess readme mckay 2025-01-26 15:17:06 +0800
  • 46eae8e02d fix refactor mckay 2025-01-26 14:05:24 +0800
  • 13617025ac director separate mckay 2025-01-26 12:52:29 +0800
  • e4b6a7b47f docs：Enhance normalization process in Source.cpp for mesh vertices mckay 2025-01-21 21:16:41 +0800
  • 8940f83fcc Add FeatureSampleConfig class for command line configuration parsing mckay 2025-01-20 23:22:59 +0800
  • a443756499 Enhance idea.md with new research points and implementation suggestions; add Dockerfile and .gitignore for pre-processing setup mckay 2025-01-18 22:06:24 +0800
  • 5f16df49e5 docs： idea.md mckay 2025-01-15 12:59:27 +0800
  • aa76cc950b Refactor evaluation.py to improve model evaluation process and error handling mckay 2025-01-05 22:50:30 +0800
  • ba5cb01580 Refactor .gitignore to streamline file exclusions and improve organization mckay 2025-01-05 22:40:05 +0800
  • f7760d5ab8 Fix path for MeshFeatureSample execution in evaluation.py to ensure correct file access during feature computation mckay 2025-01-05 22:30:14 +0800
  • 9fd733be11 Update README.md to include timezone environment variable in Docker run command mckay 2025-01-05 22:25:42 +0800
  • 0a9eb7f2ee Refactor build scripts and update CMakeLists for path consistency mckay 2025-01-05 22:24:50 +0800
  • cbf6d83529 Update .gitignore to exclude additional file types and directories mckay 2025-01-05 22:22:32 +0800
  • 5b61a72ded Update .gitignore to exclude additional files and directories mckay 2025-01-05 22:18:01 +0800
  • b2fe4ab94e Refactor evaluation.py for improved project structure and logging mckay 2025-01-05 19:06:16 +0800