diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..2df7cd6 --- /dev/null +++ b/.gitignore @@ -0,0 +1,2 @@ +.DS_store +.idea diff --git a/README.md b/README.md index edd91ed..e84d02d 100644 --- a/README.md +++ b/README.md @@ -1,2 +1,16 @@ +# Nerfies + +This is the repository that contains source code for the [Nerfies website](https://nerfies.github.io). + +If you find Nerfies useful for your work please cite: +``` +@article{park2021nerfies + author = {Park, Keunhong and Sinha, Utkarsh and Barron, Jonathan T. and Bouaziz, Sofien and Goldman, Dan B and Seitz, Steven M. and Martin-Brualla, Ricardo}, + title = {Nerfies: Deformable Neural Radiance Fields}, + journal = {ICCV}, + year = {2021}, +} +``` + # Website License Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. diff --git a/deepseekcoder.html b/deepseekcoder.html new file mode 100644 index 0000000..27ae660 --- /dev/null +++ b/deepseekcoder.html @@ -0,0 +1,341 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+
+
+
+
+

DeepSeek Coder: Empowering Everyone to Program Effortlessly

+
+ + Developed by DeepSeek AI +
+
+
+ + + + + Try Now + + + + + + + + Github + + + + + + + HuggingFace + + +
+
+
+
+ +
+
+
+
+
+
+

We introduce DeepSeek Coder, a series of large language model for coding:

+
    +
  • +     • Pretrained on 2 Trillion tokens over more than 80 programming languages. +
    • +
  • +     • Two versions(Base and Instruct) and various model sizes(1B, 7B, and 33B) to suit different needs. +
    • +
  • +     • Open source and free for research and commercial use. +
    • +
  • +     • Features a 16K window size, supporting repo-level code completion and infilling. +
    • +
  • +     • State-of-the-Art performance among open code models. The 7B base model reaches the performance of CodeLLama-34B. And the Instruction-tuned 33B model outperforms GPT3.5-turbo. +
    • +
+
+
+ 图像2 + 图像1 +
+

+ Fig. State-of-the-art Perforamnce on various Coding Benchmarks and Multilingual HumanEval +

+
+
+
+
+
+ +
+
+

Abstract

+
+

+ Deepseek Coder comprises a series of code language models trained on both 87% code and 13% natural language in English and Chinese, with each model pre-trained on 2T tokens. + We provide various sizes of the code model, ranging from 1B to 33B versions. + Each model is pre-trained on repo-level code corpus by employing a window size of 16K and a extra fill-in-the-blank task, resulting in foundational models (DeepSeek-Coder-Base). + We further fine-tune the base model with 2B tokens of instruction data to get instruction-tuned models, namedly DeepSeek-Coder-Instruct. +

+

+ For coding capabilities, DeepSeek-Coder-Base achieves state-of-the-art performance among open-source code models on multiple programming languages and various benchmarks. + And Compared to GPT35-turbo, DeepSeek-Coder-Instruct demonstrates superior performance in human evaluation while maintaining comparable performance in MBPP. +

+

+

+
+
+
+ + + +
+
+

Demo

+ Description of the GIF +

+ DeepSeek Coder empowers users to code efficiently and solve various computer problems easily +

+
+
+
+
+
+
+
+
+

Try DeepSeek Coder

+ +

+ DeepSeek Coder, powered by extensive pretraining on high-quality code data, offers accurate code completions and infills. + Its repository-level capabilities, including a 16k window size, enhance coding efficiency for real projects, making it a must-have tool(e.g., IDE Plugin) for developers. +

+ Description of the GIF +
+
+
+
+ +

DeepSeek Coder is instruction fine-tuned and aligned to help users solve various programing tasks through natural languages.

+
+
+ +
+
+
+

Code Generation

+

+ Using nerfies you can create fun visual effects. This Dolly zoom effect + would be impossible without nerfies since it would require going through a wall. +

+ +
+
+
+

Tests Generatation

+
+
+

+ As a byproduct of our method, we can also solve the matting problem by ignoring + samples that fall outside of a bounding box during rendering. +

+ Description of the GIF +
+
+
+
+
+
+
+

Demo Case 3

+

+ Using nerfies you can create fun visual effects. This Dolly zoom effect + would be impossible without nerfies since it would require going through a wall. +

+ +
+
+
+

Demo Case 4

+
+
+

+ As a byproduct of our method, we can also solve the matting problem by ignoring + samples that fall outside of a bounding box during rendering. +

+ +
+
+
+
+
+
+

Performance

+

+ We evaluate DeepSeek Coder on various coding-related benchmarks. + The result shows that DeepSeek-Coder-Base-33B significantly outperforms existing open-source code LLMs. + Compared with CodeLLama-34B, it leads by 7.9%, 9.3%, 10.8% and 5.9% respectively on HumanEval Python, HumanEval Multilingual, MBPP and DS-1000. + Surprisingly, our DeepSeek-Coder-Base-7B reaches the performance of CodeLlama-34B. + And the DeepSeek-Coder-Instruct-33B model after instruction tuning outperforms GPT35-turbo on HumanEval and achieves comparable result with GPT35-turbo on MBPP. +

+
+

+ Apart from its proficiency in coding, DeepSeek Coder also demonstrates outstanding mathematical and reasoning abilities. + Without instruction tuning, the DeepSeek-Coder-Base model has shown impressive performance in mathematical reasoning evaluations. +

+
+

Performance of different Code LLMs on various Coding Benchmarks

+ result2 +
+
+

Performance of different Open Code Models on Multilingual HumanEval

+ result_radar + +
+

Performance of different Code Models on Math-Reasoning Tasks.

+ math +
+
+ + +
+
+

How to Use DeepSeek Coder

+ +
+
+ +
+
+

Contact Us

+
+

+ If you have any questions, please raise an issue or contact us at agi_code@deepseek.com. +

+
+
+
+
+
+ + diff --git a/game.py b/game.py new file mode 100644 index 0000000..e763ea7 --- /dev/null +++ b/game.py @@ -0,0 +1,137 @@ +import pygame +import time +import random + +snake_speed = 15 + +# 窗口大小 +window_x = 720 +window_y = 480 + +# 定义颜色 +black = pygame.Color(0, 0, 0) +white = pygame.Color(255, 255, 255) +red = pygame.Color(255, 0, 0) +green = pygame.Color(0, 255, 0) +blue = pygame.Color(0, 0, 255) + +# 初始化pygame +pygame.init() + +# 初始化游戏窗口 +pygame.display.set_caption('贪吃蛇') +game_window = pygame.display.set_mode((window_x, window_y)) + +# FPS（帧率）控制器 +fps = pygame.time.Clock() + +# 蛇的初始位置 +snake_position = [100, 50] + +# 蛇的初始身体 +snake_body = [[100, 50], + [90, 50], + [80, 50], + [70, 50] + ] +# 食物的初始位置 +food_position = [random.randrange(1, (window_x//10)) * 10, + random.randrange(1, (window_y//10)) * 10] +food_spawn = True + +# 设置默认的蛇移动方向向右 +direction = 'RIGHT' +change_to = direction + +# 初始化得分 +score = 0 + +# 显示得分函数 +def show_score(choice, color, font, size): + score_font = pygame.font.SysFont(font, size) + score_surface = score_font.render('Score : ' + str(score), True, color) + score_rect = score_surface.get_rect() + game_window.blit(score_surface, score_rect) + +# 游戏结束函数 +def game_over(): + my_font = pygame.font.SysFont('times new roman', 50) + game_over_surface = my_font.render('Your Score is : ' + str(score), True, red) + game_over_rect = game_over_surface.get_rect() + game_over_rect.midtop = (window_x/2, window_y/4) + game_window.blit(game_over_surface, game_over_rect) + pygame.display.flip() + time.sleep(2) + pygame.quit() + quit() + +# 游戏主循环 +while True: + # 检查事件 + for event in pygame.event.get(): + if event.type == pygame.KEYDOWN: + if event.key == pygame.K_UP: + change_to = 'UP' + if event.key == pygame.K_DOWN: + change_to = 'DOWN' + if event.key == pygame.K_LEFT: + change_to = 'LEFT' + if event.key == pygame.K_RIGHT: + change_to = 'RIGHT' + + # 确保蛇不能反向移动 + if change_to == 'UP' and direction != 'DOWN': + direction = 'UP' + if change_to == 'DOWN' and direction != 'UP': + direction = 'DOWN' + if change_to == 'LEFT' and direction != 'RIGHT': + direction = 'LEFT' + if change_to == 'RIGHT' and direction != 'LEFT': + direction = 'RIGHT' + + # 根据方向移动蛇 + if direction == 'UP': + snake_position[1] -= 10 + if direction == 'DOWN': + snake_position[1] += 10 + if direction == 'LEFT': + snake_position[0] -= 10 + if direction == 'RIGHT': + snake_position[0] += 10 + + # 蛇吃到食物 + snake_body.insert(0, list(snake_position)) + if snake_position[0] == food_position[0] and snake_position[1] == food_position[1]: + score += 1 + food_spawn = False + else: + snake_body.pop() + + if not food_spawn: + food_position = [random.randrange(1, (window_x//10)) * 10, + random.randrange(1, (window_y//10)) * 10] + + food_spawn = True + game_window.fill(black) + + for pos in snake_body: + pygame.draw.rect(game_window, green, + pygame.Rect(pos[0], pos[1], 10, 10)) + pygame.draw.rect(game_window, white, + pygame.Rect(food_position[0], food_position[1], 10, 10)) + + # 游戏结束条件 + if snake_position[0] < 0 or snake_position[0] > window_x-10: + game_over() + if snake_position[1] < 0 or snake_position[1] > window_y-10: + game_over() + + for block in snake_body[1:]: + if snake_position[0] == block[0] and snake_position[1] == block[1]: + game_over() + + show_score(1, white, 'times new roman', 20) + + pygame.display.update() + + fps.tick(snake_speed) \ No newline at end of file diff --git a/index.html b/index.html index 0f793b1..373119f 100644 --- a/index.html +++ b/index.html @@ -2,11 +2,17 @@ - DeepSeek Coder + + + + Nerfies: Deformable Neural Radiance Fields + - - - + - - + - + @@ -69,91 +41,216 @@ -
+ + + + +
-

DeepSeek Coder:
Let the Code Write Itself

-
+

Nerfies: Deformable Neural Radiance Fields

+
+ + Keunhong Park1, + + Utkarsh Sinha2, - Developed by DeepSeek AI + Jonathan T. Barron2, + + + Sofien Bouaziz2, + + + Dan B Goldman2, + + + Steven M. Seitz1,2, + + + Ricardo Martin-Brualla2 + +
+ +
+ 1University of Washington, + 2Google Research
+
- - - Chat with DeepSeek Coder - - - - - 📑 Technical Report + + + + + Paper - - - - - Github + + + + + arXiv - - - HuggingFace + + + + + Video + - - Discord - + + + + + Code + + - - Wechat(微信) - - + + + + + Data +
+
-
+
+
+
+ +

+ Nerfies turns selfie videos from your phone into + free-viewpoint + portraits. +

+
+
+
+ + +
+
+
+ +
+
+
+ + +
@@ -161,51 +258,61 @@

Abstract

- DeepSeek Coder comprises a series of code language models trained from scratch on both 87% code and 13% natural language in English and Chinese, with each model pre-trained on 2T tokens. - We provide various sizes of the code model, ranging from 1B to 33B versions. - Each model is pre-trained on repo-level code corpus by employing a window size of 16K and a extra fill-in-the-blank task, resulting in foundational models (DeepSeek-Coder-Base). - We further fine-tune the base model with 2B tokens of instruction data to get instruction-tuned models, namedly DeepSeek-Coder-Instruct. + We present the first method capable of photorealistically reconstructing a non-rigidly + deforming scene using photos/videos captured casually from mobile phones. +

+

+ Our approach augments neural radiance fields + (NeRF) by optimizing an + additional continuous volumetric deformation field that warps each observed point into a + canonical 5D NeRF. + We observe that these NeRF-like deformation fields are prone to local minima, and + propose a coarse-to-fine optimization method for coordinate-based models that allows for + more robust optimization. + By adapting principles from geometry processing and physical simulation to NeRF-like + models, we propose an elastic regularization of the deformation field that further + improves robustness. +

+

+ We show that Nerfies can turn casually captured selfie + photos/videos into deformable NeRF + models that allow for photorealistic renderings of the subject from arbitrary + viewpoints, which we dub "nerfies". We evaluate our method by collecting data + using a + rig with two mobile phones that take time-synchronized photos, yielding train/validation + images of the same pose at different viewpoints. We show that our method faithfully + reconstructs non-rigidly deforming scenes and reproduces unseen views with high + fidelity.

- -
    -
  • - Pretrained on 2 Trillion tokens over more than 80 programming languages. -
    • -
  • - Various model sizes (1.3B, 5.7B, 6.7B and 33B) to support different requirements. -
    • -
  • - A window size of 16K window size, supporting project-level code completion and infilling. -
    • -
  • - State-of-the-Art performance among open code models. -
    • -
  • - Open source and free for research and commercial use. -
    • -
-
- Description of the GIF + + +
+
+

Video

+
+ +
+
+
+
+
- - +
-
-

Code Generation

+

Visual Effects

Using nerfies you can create fun visual effects. This Dolly zoom effect would be impossible without nerfies since it would require going through a wall. @@ -216,8 +323,11 @@

Code Generation

+ + +
-

Tests Generatation

+

Matting

@@ -229,115 +339,152 @@

Tests Generatation

type="video/mp4">
+
+ + +
-
-
-

Demo Case 3

+
+

Animation

+ + +

Interpolating states

+

- Using nerfies you can create fun visual effects. This Dolly zoom effect - would be impossible without nerfies since it would require going through a wall. + We can also animate the scene by interpolating the deformation latent codes of two input + frames. Use the slider here to linearly interpolate between the left frame and the right + frame.

-
-
-
-

Demo Case 4

-
-
-

- As a byproduct of our method, we can also solve the matting problem by ignoring - samples that fall outside of a bounding box during rendering. -

- +
+
+ Interpolate start reference image. +

Start Frame

+
+
+
+ Loading... +
+ +
+
+ Interpolation end reference image. +

End Frame

-
-
--> -
-
-

Performance

-

- We evaluate DeepSeek Coder on various coding-related benchmarks. - The result shows that DeepSeek-Coder-Base-33B significantly outperforms existing open-source code LLMs. - Compared with CodeLLama-34B, it leads by 7.9%, 9.3%, 10.8% and 5.9% respectively on HumanEval Python, HumanEval Multilingual, MBPP and DS-1000. - Surprisingly, our DeepSeek-Coder-Base-7B reaches the performance of CodeLlama-34B. - And the DeepSeek-Coder-Instruct-33B model after instruction tuning outperforms GPT-3.5-turbo on HumanEval and achieves comparable result with GPT-3.5-turbo on MBPP. -

-
-
- 图像2 - 图像1 +
+ + + +

Re-rendering the input video

+
+

+ Using Nerfies, you can re-render a video from a novel + viewpoint such as a stabilized camera by playing back the training deformations. +

-
-

- Fig. State-of-the-art Performance on various Coding Benchmarks and Multilingual HumanEval -

-
-

(1) Performance of different Code LLMs on Multilingual HumanEval Benchmark

- HumanEval -
-
-

(2) Performance of different Code LLMs on MBPP Benchmark

- MBPP -
-
-

(3) Performance of different Code LLMs on DS-1000 Benchmark

- DS-1000 -
-
-

(4) Performance of different Code Models on Math-Reasoning Tasks.

- math +
+ +
+ +
-
-
-

How to Use DeepSeek Coder

- -
-
+
-

Contact Us

+

Related Links

+

- If you have any questions, please raise an issue or contact us at agi_code@deepseek.com. + There's a lot of excellent work that was introduced around the same time as ours. +

+

+ Progressive Encoding for Neural Optimization introduces an idea similar to our windowed position encoding for coarse-to-fine optimization. +

+

+ D-NeRF and NR-NeRF + both use deformation fields to model non-rigid scenes. +

+

+ Some works model videos with a NeRF by directly modulating the density, such as Video-NeRF, NSFF, and DyNeRF +

+

+ There are probably many more by the time you are reading this. Check out Frank Dellart's survey on recent NeRF papers, and Yen-Chen Lin's curated list of NeRF papers.

+ +
+ +
+
+

BibTeX

+ 
@article{park2021nerfies,
+  author    = {Park, Keunhong and Sinha, Utkarsh and Barron, Jonathan T. and Bouaziz, Sofien and Goldman, Dan B and Seitz, Steven M. and Martin-Brualla, Ricardo},
+  title     = {Nerfies: Deformable Neural Radiance Fields},
+  journal   = {ICCV},
+  year      = {2021},
+}
+
+
+ + + diff --git a/static/.DS_Store b/static/.DS_Store deleted file mode 100644 index 037ff01..0000000 Binary files a/static/.DS_Store and /dev/null differ diff --git a/static/css/index.css b/static/css/index.css index 7b8368b..21076ef 100644 --- a/static/css/index.css +++ b/static/css/index.css @@ -1,43 +1,39 @@ body { - font-family: "Noto Sans", sans-serif; + font-family: 'Noto Sans', sans-serif; } -.hero { - padding-top: 72px; - padding-bottom: 0; -} - -.hero-body { - padding-bottom: 0; -} .footer .icon-link { - font-size: 25px; - color: #000; + font-size: 25px; + color: #000; } .link-block a { - margin-top: 5px; - margin-bottom: 5px; + margin-top: 5px; + margin-bottom: 5px; } .dnerf { font-variant: small-caps; } + .teaser .hero-body { padding-top: 0; + padding-bottom: 3rem; } .teaser { - font-family: "Google Sans", sans-serif; + font-family: 'Google Sans', sans-serif; } + .publication-title { } .publication-banner { max-height: parent; + } .publication-banner video { @@ -52,34 +48,34 @@ body { } .publication-title { - font-family: "Google Sans", sans-serif; + font-family: 'Google Sans', sans-serif; } .publication-authors { - font-family: "Google Sans", sans-serif; + font-family: 'Google Sans', sans-serif; } .publication-venue { - color: #555; - width: fit-content; - font-weight: bold; + color: #555; + width: fit-content; + font-weight: bold; } .publication-awards { - color: #ff3860; - width: fit-content; - font-weight: bolder; + color: #ff3860; + width: fit-content; + font-weight: bolder; } .publication-authors { } .publication-authors a { - color: hsl(204, 86%, 53%) !important; + color: hsl(204, 86%, 53%) !important; } .publication-authors a:hover { - text-decoration: underline; + text-decoration: underline; } .author-block { @@ -94,21 +90,21 @@ body { } .publication-video { - position: relative; - width: 100%; - height: 0; - padding-bottom: 56.25%; + position: relative; + width: 100%; + height: 0; + padding-bottom: 56.25%; - overflow: hidden; - border-radius: 10px !important; + overflow: hidden; + border-radius: 10px !important; } .publication-video iframe { - position: absolute; - top: 0; - left: 0; - width: 100%; - height: 100%; + position: absolute; + top: 0; + left: 0; + width: 100%; + height: 100%; } .publication-body img { @@ -131,6 +127,7 @@ body { margin: 0; } + .interpolation-panel { background: #f5f5f5; border-radius: 10px; @@ -158,20 +155,3 @@ body { #interpolation-image-wrapper img { border-radius: 5px; } -ul.custom-list { - list-style: none; /* 移除默认的列表样式 */ - padding-left: 0; /* 移除默认的左内边距 */ -} - -ul.custom-list li::before { - content: "•"; /* 添加自定义的项目符号 */ - color: black; /* 可以根据需要设置颜色 */ - display: inline-block; - width: 1em; /* 确保项目符号和文本之间有足够的空间 */ - margin-left: -1em; /* 将项目符号向左移动以对齐 */ -} - -ul.custom-list li { - padding-left: 1em; /* 对应于::before中的宽度，确保文本缩进 */ - text-indent: -1em; /* 当文本换行时，它将缩进以对齐到项目符号的后面 */ -} diff --git a/static/images/.DS_Store b/static/images/.DS_Store deleted file mode 100644 index 5008ddf..0000000 Binary files a/static/images/.DS_Store and /dev/null differ diff --git a/static/images/2048.gif b/static/images/2048.gif new file mode 100644 index 0000000..99f3787 Binary files /dev/null and b/static/images/2048.gif differ diff --git a/static/images/DS-1000.png b/static/images/DS-1000.png deleted file mode 100644 index dd409b0..0000000 Binary files a/static/images/DS-1000.png and /dev/null differ diff --git a/static/images/HumanEval.png b/static/images/HumanEval.png deleted file mode 100644 index a93d2c9..0000000 Binary files a/static/images/HumanEval.png and /dev/null differ diff --git a/static/images/MBPP.png b/static/images/MBPP.png deleted file mode 100644 index 427facb..0000000 Binary files a/static/images/MBPP.png and /dev/null differ diff --git a/static/images/MPBB.png b/static/images/MPBB.png deleted file mode 100644 index 1c5f79b..0000000 Binary files a/static/images/MPBB.png and /dev/null differ diff --git a/static/images/Math.png b/static/images/Math.png deleted file mode 100644 index d2592d6..0000000 Binary files a/static/images/Math.png and /dev/null differ diff --git a/static/images/code_chat.gif b/static/images/code_chat.gif deleted file mode 100644 index 26c1dee..0000000 Binary files a/static/images/code_chat.gif and /dev/null differ diff --git a/static/images/favicon.svg b/static/images/favicon.svg new file mode 100644 index 0000000..98d2546 --- /dev/null +++ b/static/images/favicon.svg @@ -0,0 +1,37 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/static/images/human-multilingual.png b/static/images/human-multilingual.png new file mode 100644 index 0000000..4776051 Binary files /dev/null and b/static/images/human-multilingual.png differ diff --git a/static/images/interpolate_end.jpg b/static/images/interpolate_end.jpg new file mode 100644 index 0000000..7792aab Binary files /dev/null and b/static/images/interpolate_end.jpg differ diff --git a/static/images/interpolate_start.jpg b/static/images/interpolate_start.jpg new file mode 100644 index 0000000..1180958 Binary files /dev/null and b/static/images/interpolate_start.jpg differ diff --git a/static/images/logo.jpeg b/static/images/logo.jpeg deleted file mode 100644 index 504a20a..0000000 Binary files a/static/images/logo.jpeg and /dev/null differ diff --git a/static/images/math.png b/static/images/math.png new file mode 100644 index 0000000..f638563 Binary files /dev/null and b/static/images/math.png differ diff --git a/static/images/merged_result.png b/static/images/merged_result.png new file mode 100644 index 0000000..22cbc98 Binary files /dev/null and b/static/images/merged_result.png differ diff --git a/static/images/result.pdf b/static/images/result.pdf deleted file mode 100644 index 901c864..0000000 Binary files a/static/images/result.pdf and /dev/null differ diff --git a/static/images/result.png b/static/images/result.png index f1b4485..e557760 100644 Binary files a/static/images/result.png and b/static/images/result.png differ diff --git a/static/images/result3.png b/static/images/result3.png index f1b4485..db783c6 100644 Binary files a/static/images/result3.png and b/static/images/result3.png differ diff --git a/static/images/steve.webm b/static/images/steve.webm new file mode 100644 index 0000000..9017a1d Binary files /dev/null and b/static/images/steve.webm differ diff --git a/static/images/table.png b/static/images/table.png index 79cf660..8d37f56 100644 Binary files a/static/images/table.png and b/static/images/table.png differ diff --git a/static/videos/.DS_Store b/static/videos/.DS_Store deleted file mode 100644 index 5008ddf..0000000 Binary files a/static/videos/.DS_Store and /dev/null differ