flagged/log.csv

질의,검색 결과,flag,username,timestamp
"몸에 반점이 있다면 어떤 질병일 가능성이 있을까?
","문서 1
정확한 원인은 밝혀져 있지 않으나 유전적, 환경적 요인이 복합되어 작용하는 것으로 짐작된다. 이 질병을 앓는 가족 구성원이 있다면 본인에게도 영향을 받을 가능성이 있다. 영향을 받은 일란성 쌍둥이가 있다면 30%의 확률로 다른 쌍둥이도 이 질병을 가지고 있을 가능성이 있다. 이 질병은 스트레스, 감염, 출생 시에 발병할 수 있다. 제1형 당뇨병, 류마티스 관절염과 같은 자가면역 질환을 가진 환자들이 영향을 받을 가능성이 있다. 흡연은 이 질병의 위험성을 높이며 안구 문제를 더 악화시킬 수 있다. 이 질병은 TSI라는 항체에서 비롯하며 이는 갑상샘 자극 호르몬과 영향도가 비슷하다. 이러한 항체들은 갑상샘이 갑상샘 호르몬을 과도하게 생산하도록 유도한다. 질병 확인을 위한 혈액 검사, 방사성 요오드 섭취를 통한 증상에 기반하여 진단한다.

문서 2
1603년 37세의 나이로 사망하였고, 아들 아흐메트 1세가 왕위를 이어받았다. 사망 원인은 정확하게 알려져 있지 않으며, 자연사나 질병일 가능성이 있다.

문서 3
이상의 연구 결과들은 오직 가스 행성 및 갈색 왜성이 존재하지 않음을 밝혀줬을 뿐 지구나 금성처럼 작은 행성이 존재할 가능성은 아직 남아 있다. 만약 뜨거운 목성이 항성 가까이 있었다면 이들은 항성 근처를 돌던 지구형 행성의 궤도를 망가뜨려 생명체 발생 가능성을 낮췄을 것이다. 따라서 가스 행성이 없다는 것은 지구 비슷한 행성이 존재할 가능성을 높여 주는 사실이 된다. 통상적으로 밝혀진 연구에 따르면 중원소 함량이 높은 별 주위에는 행성이 있을 확률이 높고 그렇지 않은 별 주위에는 행성이 있을 확률이 적은 것으로 밝혀졌다. 두꺼운 먼지 원반이 있다는 사실은 항성 가까이에 암석 행성이 존재할 가능성을 높이는 것은 사실이다. 그러나 이는 동시에 그 암석 행성이 폭격을 받을 가능성이 높음을 알려주는 사실이기도 하다. 만약 행성이 발견된다면 이후의 연구 방향은 이 행성에 생명체가 살 수 있는 물과 대기가 존재하는지를 충분한 해상도의 망원경을 통해 알아내는 것이 된다. 지구의 산소가 생명체 존재를 나타내는 척도가 되는 것처럼 무기성의 대기 조성은 원시 생명체가 존재함을 나타내는 징표가 될 수 있다.

문서 4
치매나 백혈병, 당뇨, 파킨슨병과 같은 난치병들 중에는 세포의 변이나 사멸로 인한 질병이 대다수이다. 이러한 퇴행성 질병의 경우 세포 치료법을 이용해여 치료하는 경우가 많다. 특이점주의자들에 따르면 줄기세포 연구와 같은 세포 연구는 생명 공학 연구의 일부분이며 유전자 DNA 지도를 완벽하게 구조화할 수 있다면 세포분화 치료 혹은 세포복제 치료를 통해 환자 자신의 DNA를 지니고 텔로미어가 연장된 세포를 공급할 수 있을 것이라고 본다. 예컨데 현재 당뇨병 치료에 쓰이는 거부반응 제어제가 위험한 부작용을 일으킬 가능성이 있는 반면 이러한 세포 치료는 부작용 가능성을 급격히 낮출 수 있다. 이러한 세포 치료는 단순히 난치병의 치료에만 그치는 것이 아니라, 교차분화 기술을 이용한 새로운 신체 기관으로의 교체를 가능하게 한다.

문서 5
컴퓨터에 의한 시뮬레이션에는, 보이드로 불리는 틈새나 버그가 있고, 내부로부터도 아는 경우가 있을 지도 모르다. 그러한 것을 찾아내 검증할 수 있다면, 거기에 따라 모의현실의 내부에 있는 것을 증명할 수 있을 가능성이 있다. 그러나, 물리 법칙에 반하는 일은, 그 밖에도 설명할 수 있는 가설을 생각할 수 있다(신 등). 영화 《매트릭스》로 그려진 것처럼, 기시감 등의 일상적인 기묘한 체험도 어떠한 버그로서 설명할 수 있을 가능성이 있다.

문서 6
예를 들면 대기가 없는 어떤 행성 깊은 단층 속 그늘진 곳이나 화산 동굴 속에 생태계가 형성되어 있을 가능성이 있다. 저명했던 천문학자 칼 세이건은 이 개념을 태양계에 적용하여, 1976년 논문을 통해 목성의 대기 상층부에 유기체가 살고 있을 가능성을 피력했다. 그러나 목성에는 딱딱한 표면이 없기 때문에 생명체가 존재할 가능성은 거의 없다.

문서 7
널리 알려져 있다는 사실이 반드시 참임을 보증하는 것은 아니므로 이런 주장은 논리적으로 오류이다. 개인의 신념이 잘못되어 있을 가능성이 있다면 다수의 인간의 신념도 잘못되어 있을 가능성이 있다. 예를 들면, 75%의 사람이 A라고 대답했다고 해도 25%의 확률로 A가 아닐 가능성도 있다. 이 확률이 어떻게 되든 다수가 옳다는 것은 논리적이라고 할 수 없다. 만약 만장일치라고 해도, 표본의 수가 불충분할지도 모르고, 혹은 그 사람들이 모르는 사실이 존재하고 있어서 그 사실을 알면 결과가 바뀔지도 모른다.

문서 8
""(a and b)""와 같은 논리식을 계산한다고 하면 ""a""항이 거짓인 경우에, ""b""항을 계산하지 않아도 전체 식의 답을 알 수 있다. ""(a or b)""에서 ""a""항이 참인 경우에도 마찬가지이다. 여기서 항이 복잡한 식이면 이점이 많고, 식에서 결과가 참이나 거짓일 가능성과 계산의 비용에 따라 어떤 항이 먼저 계산되어야 좋은지 알 수 있다. 따라서 ""(a or b or c)""와 같은 식에서 ""a""항이 참값을 가질 가능성이 많다면, 전체 식을 쉽게 계산할 수 있다. 이런 가능성을 보장하기 위해, 컴파일러는 더 계산해야 할 것인지, 다른 항을 지름길 계산 해야 할 것인지를 검사하기도 한다. 이런 검사는 계산을 줄이는 것을 실패할 경우나 꼭 필요한 경우 무조건 전체 식을 계산해야 할 때 시간이 더 많이 걸리게 된다.

문서 9
생물학적으로는 인간의 뇌에 필요한 유전 정보를 가지는 인공적인 게놈을 적당한 숙주의 세포에 짜넣는 것으로 인공적으로 생명을 만드는 것도 가능할지도 모른다고 생각되며, 그러한 인공생명체는 의식을 가질 가능성이 높다. 그렇지만 그 생명체 안의 어떤 속성이 의식을 낳는 것일까? 비슷한 것을 비생물학적인 부품에서 만들 수 없는 것인지? 컴퓨터를 설계하기 위한 기술로 그러한 의식체를 낳을 수 없을까? 그러한 행위는 윤리적으로 문제 없을까?와 같은 여러 문제를 안고 있다.

문서 10
3부 리그",,,2023-06-16 00:57:35.174375
"모델 경량화 및 최적화와 관련하여 Netspresso bot에게 물어보세요.
예를들어 

- Why do I need to use Netspresso?
- Summarize how to compress the model with netspresso.
- Tell me what the pruning is.
- What kinds of hardware can I use with this toolkit?
- Can I use YOLOv8 with this tool? If so, tell me the examples.","1. Why do I need to use Netspresso?
Netspresso is a machine learning pipeline that provides tools and methods for model compression and optimization. By using Netspresso, you can reduce the size and computational cost of your models, making them more efficient for deployment on resource-constrained devices. It also allows you to fine-tune and optimize your models for better performance. Overall, Netspresso helps you achieve lightweight and optimized models for real-time applications.

2. Summarize how to compress the model with Netspresso.
To compress a model with Netspresso, you can follow these steps:
- Step 1: Prepare your model.
- Step 2: Upload your model to Netspresso.
- Step 3: Choose the compression method (e.g., structured pruning, filter decomposition).
- Step 4: Configure the compression settings (e.g., pruning ratio, compression algorithm).
- Step 5: Start the compression process.
- Step 6: Check the compression result and download the compressed model.

3. Tell me what pruning is.
Pruning is a model compression technique that involves removing unnecessary parameters or structures from a complex model to make it more compact and efficient. In the context of Netspresso, pruning can be done using structured pruning, which removes entire neurons, filters, or channels from the model. This results in a simplified model that can be deployed without any specific hardware or software requirements for acceleration.

4. What kinds of hardware can I use with this toolkit?
Netspresso supports various hardware platforms for model deployment. Some of the supported hardware includes NVIDIA Jetson family devices, which can be used with TensorRT for accelerated inference. Other supported platforms include TensorFlow Lite and OpenVINO, which provide optimization for different hardware architectures. The specific hardware compatibility may depend on the compression method and the target framework used in Netspresso.

5. Can I use YOLOv8 with this tool? If so, tell me the examples.
Yes, you can use YOLOv8 with Netspresso for model compression and optimization. YOLOv8 is a popular object detection model, and Netspresso provides support for object detection models. You can follow the steps in the Netspresso documentation to prepare, upload, compress, and optimize your YOLOv8 model using the available methods and tools in Netspresso. The documentation provides examples and guidelines for using Netspresso with different models, including YOLOv8.","Source 0
Step 4: Package model (beta)
Jump to ContentDocumentationHomeDocumentationv1.3.0DocumentationLog InLog InMoon (Dark Mode)Sun (Light Mode)v1.3.0HomeDocumentationSearchINTRODUCTIONWelcomeWhat is NetsPresso?Machine Learning Pipeline with NetsPressoAdvantages of NetsPressoBENCHMARKSModel Searcher- Object DetectionModel Compressor- Image Classification- Object Detection- Semantic Segmentation- Super ResolutionUSE CASESFree People Detection ModelsFree Vehicle Detection ModelsNETSPRESSO MODEL SEARCHERFeatures & Scope of supportStep 1: Prepare datasetStep 2: Upload datasetStep 3: Create project (Quick Search)Step 3: Create project (Retraining)Step 4: Check the project result and download a modelNETSPRESSO MODEL COMPRESSORFeatures & Scope of supportMethod: Structured PruningMethod: Filter DecompositionSupported modelsSupported ONNX operatorsStep 1: Prepare modelStep 2: Upload modelStep 3: Compress model (Automatic Compression)Step 3: Compress model (Advanced Compression)Step 4: Check the compression result and download a modelStep 5: Retrain the compressed modelNETSPRESSO MODEL LAUNCHERFeatures & Scope of supportStep 1: Prepare modelStep 2: Upload modelStep 3: Convert model (beta)Step 4: Package model (beta)Personal serverRequirementsConnect Personal ServerRELEASE NOTESNetsPresso ReleasesFAQAbout Credit?Powered by Step 4: Package model (beta)Suggest EditsPackage the model so that it can be deployed directly to the device. 
1. Go to Package page
Select New package at the drop-down menu that appears when you click Download button.
2. Package the model
Enter the package name and select a base model to package.
Please note that the package name will be the library name and the name cannot be changed after packaging.
You can include pre/post processing codes (.py) with the model for the package (optional).
Download the pre/post processing code example and modify for your use cases.
3. Download package file and run the package
Packaged file will be automatically downloaded.
To run the package, use the code written below. {package_name} must be changed to your package name.
Pythonfrom np_{package_name}.models.model import NPModel
NPModel.initialize(num_threads=1) # Initialize
npmodel = NPModel()
image_path = ""./test.jpg"" #Image path
print(npmodel.run(image_path)) # Inference
NPModel.finalize() # Memory management
Updated about 1 month ago Table of Contents
1. Go to Package page
2. Package the model
3. Download package file and run the package

######################################################

Source 1
Step 3: Convert model (beta)
Jump to ContentDocumentationHomeDocumentationv1.3.0DocumentationLog InLog InMoon (Dark Mode)Sun (Light Mode)v1.3.0HomeDocumentationSearchINTRODUCTIONWelcomeWhat is NetsPresso?Machine Learning Pipeline with NetsPressoAdvantages of NetsPressoBENCHMARKSModel Searcher- Object DetectionModel Compressor- Image Classification- Object Detection- Semantic Segmentation- Super ResolutionUSE CASESFree People Detection ModelsFree Vehicle Detection ModelsNETSPRESSO MODEL SEARCHERFeatures & Scope of supportStep 1: Prepare datasetStep 2: Upload datasetStep 3: Create project (Quick Search)Step 3: Create project (Retraining)Step 4: Check the project result and download a modelNETSPRESSO MODEL COMPRESSORFeatures & Scope of supportMethod: Structured PruningMethod: Filter DecompositionSupported modelsSupported ONNX operatorsStep 1: Prepare modelStep 2: Upload modelStep 3: Compress model (Automatic Compression)Step 3: Compress model (Advanced Compression)Step 4: Check the compression result and download a modelStep 5: Retrain the compressed modelNETSPRESSO MODEL LAUNCHERFeatures & Scope of supportStep 1: Prepare modelStep 2: Upload modelStep 3: Convert model (beta)Step 4: Package model (beta)Personal serverRequirementsConnect Personal ServerRELEASE NOTESNetsPresso ReleasesFAQAbout Credit?Powered by Step 3: Convert model (beta)Suggest EditsConvert the model format according to the target device.
1. Go to Convert page
Click the Convert button on Models page.
2. Covert model
Enter the name and memo for the converted model. Select a base model to be converted and the target hardware to benchmark the model.
Depending on the framework of the base model, the options available for converting are different.
Models built with Model Searcher → TensorRT, TensorFlow Lite, OpenVINO
Custom models
ONNX → TensorRT, TensorFlow Lite, OpenVINO
Click the Start converting button to convert the model. (Converting for the NVIDIA Jetson family (TensorRT) may take up to 1 hour.)
3. Check the converting result
Converted model will be displayed on the Models page with performance benchmarks on the selected target hardware.Updated 6 months ago Table of Contents
1. Go to Convert page
2. Covert model
3. Check the converting result

######################################################

Source 2
Method: Structured Pruning
Jump to ContentDocumentationHomeDocumentationv1.3.0DocumentationLog InLog InMoon (Dark Mode)Sun (Light Mode)v1.3.0HomeDocumentationSearchINTRODUCTIONWelcomeWhat is NetsPresso?Machine Learning Pipeline with NetsPressoAdvantages of NetsPressoBENCHMARKSModel Searcher- Object DetectionModel Compressor- Image Classification- Object Detection- Semantic Segmentation- Super ResolutionUSE CASESFree People Detection ModelsFree Vehicle Detection ModelsNETSPRESSO MODEL SEARCHERFeatures & Scope of supportStep 1: Prepare datasetStep 2: Upload datasetStep 3: Create project (Quick Search)Step 3: Create project (Retraining)Step 4: Check the project result and download a modelNETSPRESSO MODEL COMPRESSORFeatures & Scope of supportMethod: Structured PruningMethod: Filter DecompositionSupported modelsSupported ONNX operatorsStep 1: Prepare modelStep 2: Upload modelStep 3: Compress model (Automatic Compression)Step 3: Compress model (Advanced Compression)Step 4: Check the compression result and download a modelStep 5: Retrain the compressed modelNETSPRESSO MODEL LAUNCHERFeatures & Scope of supportStep 1: Prepare modelStep 2: Upload modelStep 3: Convert model (beta)Step 4: Package model (beta)Personal serverRequirementsConnect Personal ServerRELEASE NOTESNetsPresso ReleasesFAQAbout Credit?Powered by Method: Structured PruningSuggest EditsModel Compression
The goal of model compression is to achieve a model that is simplified from the original without performance deterioration. By compressing the large model, the user can reduce the storage and computational cost and allow to use in real-time applications.
NetsPresso supports the following compression methods.
Structured Pruning
Filter Decomposition
This page describes for Structured Pruning.
What is ""Pruning""?
	Pruning is the process of removing individual or groups of parameters from a complex model to make it faster and more compact. This compressing procedure is divided into unstructured pruning and structured pruning by the pruning objects.
Unstructured Pruning
    
			: Removes individual parameters and returns a sparse model, which requires an additional device to be accelerated.
    
Structured Pruning
    
			: Removes entire neurons, filters, or channels and returns a model, which does not require any particular hardware or software to be accelerated.
    
  The goal of pruning is to reduce the computational resources and accelerate the model by removing unnecessary filters (Model Compressor only supports structured pruning. Unstructured pruning will be published in near future.).
    However, the fine-tuning process is necessary to compensate for the loss of accuracy.
Structured Pruning
Supported functions
  Pruning in Model Compressor provides two pruning functions (Pruning by Channel Index / Criteria) and one recommendation (SLAMP) to fulfill the user's demand on model compression.
Pruning by Channel Index
    
  Removes the filters that a user wants to. If the selected filters are redundant or less important, it will return a better performing model.
    
 Pruning by Criteria
  	
L2 Norm
	      
          : L2-Norm is used to represent the importance of the corresponding filter. In other words, this method prunes filters based on the magnitude of weights.
        
Nuclear Norm 
	      
           : The nuclear norm is the sum of the singular values representing the energy. It computes the nuclear norm on the feature map to determine the filter's relevance. For this reason, a portion of the dataset is needed. For more detail, please refer to the following paper.
        
Seul-Ki Yeom, Kyung-Hwan Shim, and Jee-Hyun Hwang. Toward compact deep neural networks via energy-aware pruning. arXiv preprint, 2021.
              
Geometric Median
        
        	: Geometric Median is used to measure the redundancy of the corresponding filter and remove redundant filters. For more detail, please refer to the following paper.
        
Yang He, Ping Liu, Ziwei Wang, Zhilan Hu, and Yi Yang. Filter pruning via geometric median for deep convolutional neural networks acceleration. In CVPR, 2019.
              
Normalization
        
The distribution and magnitude of the layers are varied, it is vital to compare those different distributions from the same perspective. For this reason, all of the criterion values are normalized by layer as follows.
            
 ""Recommendation"" in Model Compressor
    
  The ""Recommendation"" enables a so-called global pruning, which allocates the pruning ratio for each layer at ease. Current version only supports SLAMP.
    
SLAMP (Structured Layer-adaptive Sparsity for the Magnitude-based Pruning)
      
SLAMP is inspired by the ""Layer-adaptive Sparsity for the Magnitude-based Pruning"" from ICLR 2021, which is called LAMP.
        
Layer-Adaptive sparsity for the Magnitude-based Pruning (LAMP) is an unstructured pruning method, but here, it is modified and developed to measure the layer-wise importance for the Structured pruning.
        
Normalization function
          
 Following normalization function is adopted into the above criteria value.
          
What you can do with Model Compressor
 Choose one of ""Pruning by Channel Index"" or ""Pruning by Criteria"" by the purpose of the user.
    
 ""Pruning by Channel Index"" is recommended for the expert, who already knows which filter is unnecessary to solve the issue.
       ""Pruning by Criteria"" is recommended for the user who wants to prune certain amounts of ratio on specific layers or beginner of the model compression.
    
To use ""Pruning by Channel Index""
 Check ""Pruning by Channel Index"".
         Check the layers to be pruned.
         Insert the filter index to be pruned (ex. 105, 8, 9, 11-13).
    
 To use ""Pruning by Criteria"".
    
 Check one of the criteria (ex. L2-Norm) to calculate the importance score of each filter.
       Prune specific layers with a certain amount of ratio.
        
 Check the layers to be pruned.
           Insert the amounts of the ratio to be pruned (ex. 0.2).
        
 Compress whole layers automatically for the given amounts of ratio.
        
 Press ""Recommendation"" and insert the amounts of the ratio (ex. 0.8) to compress the model.
        
Verification of the Structured Pruning
  Pruning in Model Compressor allows to remove the specific filters and preserve the previous knowledge. To verify the pruning method in Model Compressor, we assess the categorical accuracy of the returned compressed model and its fine-tuned model.
  
 Cifar100 Accuracy of the Compressed Model and Fine-tuned Model
The compressed models are compressed by the following strategies: L2Norm, SLAMP, intersection. (Original models are from  here)
The dashed and solid lines in the above image indicate the compressed model from Model Compressor and the model after fine-tuning, respectively, for the given pruning ratio.
🚧Not Supported Layers
Group convolutional layer is currently not supported and will be updated in the near future.
Updated 5 months ago Table of Contents
Model Compression
What is ""Pruning""?
Structured Pruning
Supported functions
What you can do with Model Compressor
Verification of the Structured Pruning
",,,2023-06-28 07:12:14.309344
What is NetsPresoo?,"NetsPresso is a machine learning pipeline that provides various tools and functionalities for model development and optimization. It includes components such as Model Searcher, Model Compressor, and Model Launcher. NetsPresso aims to simplify the process of developing and deploying machine learning models by offering features like dataset preparation, model compression, and model deployment. It supports tasks like image classification, object detection, semantic segmentation, and super resolution. NetsPresso also provides benchmark results and best practices for different models and datasets.","Source 0
Model Compressor
Jump to ContentDocumentationHomeDocumentationv1.3.0DocumentationLog InLog InMoon (Dark Mode)Sun (Light Mode)v1.3.0HomeDocumentationSearchINTRODUCTIONWelcomeWhat is NetsPresso?Machine Learning Pipeline with NetsPressoAdvantages of NetsPressoBENCHMARKSModel Searcher- Object DetectionModel Compressor- Image Classification- Object Detection- Semantic Segmentation- Super ResolutionUSE CASESFree People Detection ModelsFree Vehicle Detection ModelsNETSPRESSO MODEL SEARCHERFeatures & Scope of supportStep 1: Prepare datasetStep 2: Upload datasetStep 3: Create project (Quick Search)Step 3: Create project (Retraining)Step 4: Check the project result and download a modelNETSPRESSO MODEL COMPRESSORFeatures & Scope of supportMethod: Structured PruningMethod: Filter DecompositionSupported modelsSupported ONNX operatorsStep 1: Prepare modelStep 2: Upload modelStep 3: Compress model (Automatic Compression)Step 3: Compress model (Advanced Compression)Step 4: Check the compression result and download a modelStep 5: Retrain the compressed modelNETSPRESSO MODEL LAUNCHERFeatures & Scope of supportStep 1: Prepare modelStep 2: Upload modelStep 3: Convert model (beta)Step 4: Package model (beta)Personal serverRequirementsConnect Personal ServerRELEASE NOTESNetsPresso ReleasesFAQAbout Credit?Powered by Model CompressorSuggest EditsUsing NetsPresso Model Compressor on a variety of well-known public datasets, this article shows users how to get outstanding compression results.
The fine-tuning procedure is necessary for each compression. It usually follows the original model's training configuration, except the learning rate. After a few batches of training, the learning rate is optimized by determining if the loss has converged or not.
All of the original and compressed models can be downloaded easily on the Model Compressor Model Zoo.
See Image Classification Results
See Object Detection Results
See Semantic Segmentation Results
See Super Resolution ResultsUpdated 7 months ago Table of Contents
See Image Classification Results
See Object Detection Results
See Semantic Segmentation Results
See Super Resolution Results

######################################################

Source 1
Connect Personal Server
Jump to ContentDocumentationHomeDocumentationv1.3.0DocumentationLog InLog InMoon (Dark Mode)Sun (Light Mode)v1.3.0HomeDocumentationSearchINTRODUCTIONWelcomeWhat is NetsPresso?Machine Learning Pipeline with NetsPressoAdvantages of NetsPressoBENCHMARKSModel Searcher- Object DetectionModel Compressor- Image Classification- Object Detection- Semantic Segmentation- Super ResolutionUSE CASESFree People Detection ModelsFree Vehicle Detection ModelsNETSPRESSO MODEL SEARCHERFeatures & Scope of supportStep 1: Prepare datasetStep 2: Upload datasetStep 3: Create project (Quick Search)Step 3: Create project (Retraining)Step 4: Check the project result and download a modelNETSPRESSO MODEL COMPRESSORFeatures & Scope of supportMethod: Structured PruningMethod: Filter DecompositionSupported modelsSupported ONNX operatorsStep 1: Prepare modelStep 2: Upload modelStep 3: Compress model (Automatic Compression)Step 3: Compress model (Advanced Compression)Step 4: Check the compression result and download a modelStep 5: Retrain the compressed modelNETSPRESSO MODEL LAUNCHERFeatures & Scope of supportStep 1: Prepare modelStep 2: Upload modelStep 3: Convert model (beta)Step 4: Package model (beta)Personal serverRequirementsConnect Personal ServerRELEASE NOTESNetsPresso ReleasesFAQAbout Credit?Powered by Connect Personal ServerSuggest EditsTo connect a personal server, start by clicking a 'New Server' button on the screen below.
location : My Account > Resources
The way to connect a personal server to NetsPresso is to install an agent on the personal server.
The process is as follows.
Step 1. Create Server
Specifies an identifiable name and the path where the agent will be installed.
Step 2. Set Server
Copy the script that pre-checks the server environment and receives server information.
Make the copied script into sh file and run it in the shell.
As a result of executing the script, you can see server information in json format as shown below.
Copy the server information in json format and paste it into the result input field.
Step 3. Connect Server
As shown below, check the server information and copy the connect script that can connect the server.
Make the copied script into sh file and run it in the shell.
As a result of execution, the server is connected as shown below.
You can check the server you have connected to on the Resources page.
Updated about 1 month ago Table of Contents
Step 1. Create Server
Step 2. Set Server
Step 3. Connect Server

######################################################

Source 2
- Object Detection
Jump to ContentDocumentationHomeDocumentationv1.3.0DocumentationLog InLog InMoon (Dark Mode)Sun (Light Mode)v1.3.0HomeDocumentationSearchINTRODUCTIONWelcomeWhat is NetsPresso?Machine Learning Pipeline with NetsPressoAdvantages of NetsPressoBENCHMARKSModel Searcher- Object DetectionModel Compressor- Image Classification- Object Detection- Semantic Segmentation- Super ResolutionUSE CASESFree People Detection ModelsFree Vehicle Detection ModelsNETSPRESSO MODEL SEARCHERFeatures & Scope of supportStep 1: Prepare datasetStep 2: Upload datasetStep 3: Create project (Quick Search)Step 3: Create project (Retraining)Step 4: Check the project result and download a modelNETSPRESSO MODEL COMPRESSORFeatures & Scope of supportMethod: Structured PruningMethod: Filter DecompositionSupported modelsSupported ONNX operatorsStep 1: Prepare modelStep 2: Upload modelStep 3: Compress model (Automatic Compression)Step 3: Compress model (Advanced Compression)Step 4: Check the compression result and download a modelStep 5: Retrain the compressed modelNETSPRESSO MODEL LAUNCHERFeatures & Scope of supportStep 1: Prepare modelStep 2: Upload modelStep 3: Convert model (beta)Step 4: Package model (beta)Personal serverRequirementsConnect Personal ServerRELEASE NOTESNetsPresso ReleasesFAQAbout Credit?Powered by - Object DetectionSuggest EditsAll of the original and compressed models can be downloaded easily on the Model Compressor Model Zoo.
You can get Compressed results with Automatic Compression and Compressed (Adv.) results with Advanced Compression.
PyTorch
ModelBest PracticeTypeDatasetmAP(0.5) (%)mAP(0.5:0.95)(%)FLOPs (M)Params (M)Latency (ms)Model Size (MB)YOLOXOriginalCOCO68.049.7156006.2054.2112239.46207.37YOLOXGoogle ColabCompressed-1COCO67.16 (-0.84)48.64 (-1.06)101804.06 (1.53x)19.96 (2.7x)8502.72 (1.44x)76.61 (2.7x)YOLOXGoogle ColabCompressed-2COCO61.43 (-6.57)43.23 (-5.47)38607.03 (4.04x)4.93 (11.0x)4235.37 (2.89x)19.17 (10.80x)
The model’s latency is measured on Raspberry Pi 4B (1.5GHz ARM Cortex).
Options: FP32, ONNX runtime
TensorFlow-Keras
ModelBest PracticeTypeDatasetmAP(0.5) (%)mAP(0.5:0.95)(%)FLOPs (M)Params (M)Latency (ms)Model Size (MB)YOLOv4OriginalPASCAL VOC82.22-61871.8265.3264318.70262.90YOLOv4Google ColabCompressed-1PASCAL VOC87.23 (+5.01)-11459.69 (5.4x)10.59 (6.17x)28651.70 (2.16x)44.12 (5.96x)YOLOv4Google ColabCompressed-2PASCAL VOC87.91 (+5.69)-14442.96 (4.28x)10.71 (6.1x)28976.40 (2.14x)44.36 (5.93x)
YOLOv4 model with EfficientNet B1 based backbone.
The model’s latency is measured on Raspberry Pi 4B (1.5GHz ARM Cortex).
Options: FP32, TFLite
Updated about 2 months ago Table of Contents
PyTorch
TensorFlow-Keras

######################################################

Source 3
- Object Detection
Jump to ContentDocumentationHomeDocumentationv1.3.0DocumentationLog InLog InMoon (Dark Mode)Sun (Light Mode)v1.3.0HomeDocumentationSearchINTRODUCTIONWelcomeWhat is NetsPresso?Machine Learning Pipeline with NetsPressoAdvantages of NetsPressoBENCHMARKSModel Searcher- Object DetectionModel Compressor- Image Classification- Object Detection- Semantic Segmentation- Super ResolutionUSE CASESFree People Detection ModelsFree Vehicle Detection ModelsNETSPRESSO MODEL SEARCHERFeatures & Scope of supportStep 1: Prepare datasetStep 2: Upload datasetStep 3: Create project (Quick Search)Step 3: Create project (Retraining)Step 4: Check the project result and download a modelNETSPRESSO MODEL COMPRESSORFeatures & Scope of supportMethod: Structured PruningMethod: Filter DecompositionSupported modelsSupported ONNX operatorsStep 1: Prepare modelStep 2: Upload modelStep 3: Compress model (Automatic Compression)Step 3: Compress model (Advanced Compression)Step 4: Check the compression result and download a modelStep 5: Retrain the compressed modelNETSPRESSO MODEL LAUNCHERFeatures & Scope of supportStep 1: Prepare modelStep 2: Upload modelStep 3: Convert model (beta)Step 4: Package model (beta)Personal serverRequirementsConnect Personal ServerRELEASE NOTESNetsPresso ReleasesFAQAbout Credit?Powered by - Object DetectionSuggest EditsQuick Search supports to train a model based on open sources and NPNets will be available with Advanced Search.
YOLOv5 and YOLOv5-NPNets
DatasetModelmAP(0.5) (%)mAP(0.5:0.95) (%)GFLOPsParameters (M)Model Size (MB)PASCAL VOCYOLOv5n72.2444.31.793.62PASCAL VOCYOLOv5n-NPNet73.446.33.51.182.49PASCAL VOCYOLOv5s77.952.2167.0713.7PASCAL VOCYOLOv5s-NPNet80.25612.84.619.05PASCAL VOCYOLOv5m82.159.348.320.9540.2PASCAL VOCYOLOv5m-NPNet83.460.63712.2623.7PASCAL VOCYOLOv5l82.961108.346.2488.6PASCAL VOCYOLOv5l-NPNet85.163.88125.1948.5Updated 5 months ago Table of Contents
YOLOv5 and YOLOv5-NPNets
",,,2023-08-31 13:29:07.271798