Batch Inference to get bounding box coordinates using TensorRT

[santoshmedisetty]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

Hi, I checked the issues and discussions and couldn't find a proper resource. I converted my custom yolov5n .pt model to .engine model using export.py. Now I want to perform batch inference on images to get the bounding box coordinates. Can anyone please help me or guide me to a resource?

Additional

No response

[github-actions]

👋 Hello @santoshmedisetty, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://ultralytics.com or email Glenn Jocher at glenn.jocher@ultralytics.com.

Requirements

Python>=3.6.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab and Kaggle notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training (train.py), validation (val.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

See export.py argparser arguments for exporting at arbitrary batch sizes, i.e. --batch 64.

!pip install -U nvidia-tensorrt --index-url https://pypi.ngc.nvidia.com  # install
!python export.py --weights yolov5s.pt --include engine --imgsz 640 640 --device 0  # export
!python detect.py --weights yolov5s.engine --imgsz 640 640 --device 0  # inference

[santoshmedisetty]

Thanks @glenn-jocher for the prompt reply. I exported the model to .engine by specifying the required batch size. I was looking for batch inference solution similar to #1806 to get the bounding box coordinates.

[glenn-jocher]

@santoshmedisetty you can use your .engine model in detect.py, but this method usually uses batch-size 1 dataloaders. If you load your engine model with PyTorch Hub you can use any batch size:

YOLOv5 🚀 PyTorch Hub models allow for simple model loading and inference in a pure python environment without using detect.py.

Simple Inference Example

This example loads a pretrained YOLOv5s model from PyTorch Hub as model and passes an image for inference. 'yolov5s' is the lightest and fastest YOLOv5 model. For details on all available models please see the README.

import torch

# Model
model = torch.hub.load('ultralytics/yolov5', 'yolov5s')  # or yolov5m, yolov5l, yolov5x, custom

# Images
im = 'https://ultralytics.com/images/zidane.jpg'  # or file, Path, URL, PIL, OpenCV, numpy, list

# Inference
results = model(im)

# Results
results.print  # or .show(), .save(), .crop(), .pandas(), etc.

results.xyxy[0]  # im predictions (tensor)
results.pandas().xyxy[0]  # im predictions (pandas)
#      xmin    ymin    xmax   ymax  confidence  class    name
# 0  749.50   43.50  1148.0  704.5    0.874023      0  person
# 2  114.75  195.75  1095.0  708.0    0.624512      0  person
# 3  986.00  304.00  1028.0  420.0    0.286865     27     tie

See YOLOv5 PyTorch Hub Tutorial for details.

Good luck 🍀 and let us know if you have any other questions!

[santoshmedisetty]

Hi @glenn-jocher, First of all, a big thank you for helping me. I loaded my .engine model using pytorch hub but got an error while making predictions(result = model(inputs)). Both my inputs and the model are on the same device(GPU). Please see the below error message.

RuntimeError: CUDA error: an illegal memory access was encountered
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
[01/10/2022-19:04:40] [TRT] [E] 1: [defaultAllocator.cpp::deallocate::35] Error Code 1: Cuda Runtime (an illegal memory access was encountered)
[01/10/2022-19:04:40] [TRT] [E] 1: [defaultAllocator.cpp::deallocate::35] Error Code 1: Cuda Runtime (an illegal memory access was encountered)

P.S: I used 'torch.cuda.current_device()' at the beginning of my code

[glenn-jocher]

@santoshmedisetty I'll take a look. I haven't actually tried to run TRT via torch hub models yet.

[glenn-jocher]

@santoshmedisetty I'm able to reproduce an error in Colab using this code (after running Setup cell):

!pip install -U nvidia-tensorrt --index-url https://pypi.ngc.nvidia.com  # install
!python export.py --weights yolov5s.pt --include engine --imgsz 640 640 --device 0  # export
!python detect.py --weights yolov5s.engine --imgsz 640 640 --device 0  # inference

# PyTorch Hub
import torch
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.engine')
model('https://ultralytics.com/images/zidane.jpg').print()

I'll add a TODO here to look into this but I can't provide a timeline for a fix currently. I've verified detect.py inference works correctly (per reproduction code above).

TODO: TRT PyTorch Hub inference fix

[santoshmedisetty]

Thank you @glenn-jocher. I'll wait for the fix.

[glenn-jocher]

@santoshmedisetty the issue seems to originate with a missing LD_LIBRARY_PATH variable, i.e. depending on your installation:

%env LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64:/content/cuda-11.1/lib64:/content/TensorRT-8.2.0.6/lib:/usr/local/nvidia/lib:/usr/local/nvidia/lib64  # add to path

See https://forums.developer.nvidia.com/t/importerror-libnvinfer-so-4-cannot-open-shared-object-file-no-such-file-or-directory/62243/4

[glenn-jocher]

@santoshmedisetty the error is reproducible very simple with:

import tensorrt

[glenn-jocher]

@santoshmedisetty ok so it seems the above info only holds for the tar install. For the new pip install instructions there is a different error only on hub inference (detect.py is fine):

# TensorRT 
# https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html#installing-pip
!pip install -U nvidia-tensorrt --index-url https://pypi.ngc.nvidia.com  # install
!python export.py --weights yolov5s.pt --include engine --imgsz 640 640 --device 0  # export
!python detect.py --weights yolov5s.engine --imgsz 640 640 --device 0  # inference

# PyTorch Hub
import torch
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.engine')
model('https://ultralytics.com/images/zidane.jpg').print()

[santoshmedisetty]

@glenn-jocher So, do you mean, I'm getting the error because of TensorRT installation? I'm running my code on Ubuntu 20.10. And installed tensorrt using pip

[glenn-jocher]

@santoshmedisetty error with pip install is

RuntimeError: CUDA error: an illegal memory access was encountered
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.