gengyanlei / fire-smoke-detect-yolov4 Goto Github PK

您好，下载的数据中没有提供烟雾这个类别。请问是我解析错了吗？还是就是没有提供

这是两者的inference 输出 :
darknet box: [0.44398072, 0.25001103, 0.10880162 ,0.27873766] conf : 0.6275723
onnx box: [0.3904399 ,0.10827315 ,0.49689564 ,0.39131567] conf: 0.67223823

这是我用的测试的图片：
https://image.baidu.com/search/detail?ct=503316480&z=undefined&tn=baiduimagedetail&ipn=d&word=%E6%89%93%E7%81%AB%E6%9C%BA&step_word=&ie=utf-8&in=&cl=2&lm=-1&st=undefined&hd=undefined&latest=undefined&copyright=undefined&cs=2444222964,2620753514&os=2385169486,2135497603&simid=3447159448,455013216&pn=2&rn=1&di=91520&ln=1681&fr=&fmq=1609410555919_R&fm=&ic=undefined&s=undefined&se=&sme=&tab=0&width=undefined&height=undefined&face=undefined&is=0,0&istype=0&ist=&jit=&bdtype=0&spn=0&pi=0&gsm=0&objurl=https%3A%2F%2Fgimg2.baidu.com%2Fimage_search%2Fsrc%3Dhttp%253A%252F%252Fimg.alicdn.com%252Fbao%252Fuploaded%252Fi4%252F2682881798%252FTB2ef_WfmtYBeNjSspaXXaOOFXa_%2521%25212682881798.jpg_300x300.jpg%26refer%3Dhttp%253A%252F%252Fimg.alicdn.com%26app%3D2002%26size%3Df9999%2C10000%26q%3Da80%26n%3D0%26g%3D0n%26fmt%3Djpeg%3Fsec%3D1612002549%26t%3Ddee2278454e7ce26bfb4b948c703392d&rpstart=0&rpnum=0&adpicid=0&force=undefined

@gengyanlei 请问下是什么原因呢？谢谢

I cannot access the fire-yolov4-weights and fire-dataset from the cloud Baidu... Could you please upload them on google drive or similar ? Thanks

I am using opencv 4.5.4, which supports the use of converted yolov5.onnx. So, I converted the best.pt (inside Yolov5 folder) to best.onnx.

I can now load the model successfully using: cv::dnn::readNetFromONNX("best.onnx")
BUT, I do not know how to use it to find bounding boxes around fires and smokes given a test image!

Please help me anyone does know how. thanks.

我解压后，为什么这个文件没有后缀名呢， 没法打开，请问能帮忙解决一下吗？

Hi,
thanks for sharing your work. Are there some restrictions in using the dataset that you provide?
What is the license ?

Thanks for your help

如题

@ALL
【1】此权重为yolov5 v3版本的代码，未做修改，有些按照我的readme执行代码，并未产生结果，未果不清楚具体原因
【2】你们可以直接下载yolov5 v3版本代码，然后load我的烟火2类权重即可，检测效果虽然不是最好的，但还可以！

你好呀，我通过你的下载链接下载数据集后，里面的文件只有火焰的图片，请问哪个链接是和你用yolov5训练的火焰+烟雾的数据呢？

fire.names里面只有火焰，经测试只能识别火焰，不能识别烟雾，请问用Yolov4算法如何识别烟雾？

我最近也在做yolov4的火焰检测，能加QQ好友吗，接下来准备做yolov4剪枝和tensorrt部署。QQ：406174050
我因为工作可以去工厂采集数据，后面也方便建立数据集

我将下载好的文件放到树莓派去运行，但是出现这个问题，请问是因为CUDA没安装的原因吗？我树莓派的yolov5环境已经搭建好的。

Excuse me, i already train my weight using yoloV4 configuration from alexey and some augmentation data like rotate, brightness contrast, etc and have running about 6000 iteration but only got Map around 65% when using your dataset , when i see your weight you have Map around 75% could you please tell me what data augmentation are you using or what hyperparameter and augmentation best for this case , thanks

下载了您提供的 “烟火(2059张图像,含标签)” 后发现里面只有含有火焰的标签，并未发现烟雾的标签。

我想使用您的YOLO v5 实现一个能同时检测烟雾和火焰的检测器，请问您用来训练v5的数据集烟雾火焰数据集可以开源吗？谢谢！

I used weight file on yolov5: best.pt. I see it is good. Can you tell me how many epoch did you train ? And Can you share me dataset about fire-smoke?

可参考：
https://github.com/gengyanlei/onnx2tensorRt

在训练yolov5的时候遇到这个错误，请问这个要怎么解决？

提取码：hhwq

此外：
感谢分享数据的小伙伴，由于涉及小伙伴的成果，因此剩余的图像标注不开源！

【1】烟雾-进一步细分成 白烟-黑烟-灰烟 or 浓烟 淡烟
【2】火 也可以进一步细分，因为这会增加网络训练难度，但是可以提升识别效果

(next mAP calculation at 6175 iterations)
Last accuracy [email protected] = 62.27 %, best = 66.48 %
6000: 1.022357, 1.471952 avg loss, 0.000010 rate, 4.679000 seconds, 192000 images, 0.091826 hours left
Resizing to initial size: 416 x 416 try to allocate additional workspace_size = 52.43 MB
CUDA allocate done!

calculation mAP (mean average precision)...
Detection layer: 139 - type = 28
Detection layer: 150 - type = 28
Detection layer: 161 - type = 28
260
detections_count = 1069, unique_truth_count = 413
class_id = 0, name = fire, ap = 61.61% (TP = 258, FP = 106)

for conf_thresh = 0.25, precision = 0.71, recall = 0.62, F1-score = 0.66
for conf_thresh = 0.25, TP = 258, FP = 106, FN = 155, average IoU = 54.18 %

IoU threshold = 50 %, used Area-Under-Curve for each unique Recall
mean average precision ([email protected]) = 0.616056, or 61.61 %
Total Detection Time: 9 Seconds

感觉指标是不是有点低啊=-=

Please give me yolov4fire-best.weights.
Thank you so much

FileNotFoundError: Could not find module 'D:\fire_smoke_detect\fire-smoke-detect-yolov4-master\yolov4\yolo_cpp_dll.dll'. Try using the full path with constructor syntax.
我在GitHub上找了个yolo_cpp_dll.dll放到工程中去但是还是报这个错误

实际预训练的模型，能否检测烟雾呢？因为我看到demo图片有一张是有烟雾检测的分类的。

您好，请问这个数据集：烟火(2059张图像,含标签)-百度云盘下载链接 提取码->(3q4r) 我下载解压后为什么是一个整体文件呢，不是pascol类型的数据集，是还需要做其他操作吗？

当我调用latest_darknet_API.py进行推理时，报如下错误：OSError: libopencv_core.so.2.4: cannot open shared object file: No such file or directory.请问怎么解决

频繁调用detect.py函数，若检测到目标会导致内存爆增300mb左右

Can you share weights file to detect smoke and fire.
Thank you!

可以在arm架构上跑吗

请问是使用开源的yolov5权重对未标注图像进行预测，然后对预测结果进行标注吗？数据集的标注没有使用labeling吗？

你好，十分感谢你的开源项目，这是一个非常意思的工作。当我尝试在我的系统下运行你的代码时，遇到了如title的错误，我想请假下你，这是由于什么原因造成的？是因为我的opencv版本不对吗？我的opencv版本是4.3的，期待你的解答，谢谢。

it is required in darknet.py

best.pt 这是训练后的模型么？DLJ调用yolov5模型遇到一些问题

array_to_image函数不存在

Unable to download the yolov4 weights because this link is not opening.

[yolo] params: iou loss: ciou (4), iou_norm: 0.07, obj_norm: 1.00, cls_norm: 1.00, delta_norm: 1.00, scale_x_y: 1.05
nms_kind: greedynms (1), beta = 0.600000
Total BFLOPS 127.232
avg_outputs = 1046213
Allocate additional workspace_size = 52.43 MB
Try to load weights: ./backup_fire/yolov4-fire_best.weights
Loading weights from ./backup_fire/yolov4-fire_best.weights...
seen 64, trained: 128 K-images (2 Kilo-batches_64)
Done! Loaded 162 layers from weights-file
Loaded - names_list: ./cfg/fire.names, classes = 1
Segmentation fault (core dumped)
只要调用predict_image就会崩溃，而且没有任何有效信息。

注释：
yolov5s 烟火检测模型是我随便训练几个epoch的结果，并不是最好的结果。目前只能开源1个看起来还说的过去的模型。最好的模型不会开源的。

Hello, Thank you for your great work
Can you provide a Colab Notebook Demo for this amazing repo, It will help everyone and it will be much appreciated,
thank you

您的資料集內只有火焰的標記，可否請大佬提供包含火焰集煙霧的資料集？謝謝。

执行 apt-get install libopencv-dev之后安装了最新版本，提示”libopencv-dev is already the newest version (3.2.0+dfsg-4ubuntu0.1).“

但该问题任然存在

dear team i am a person who don't have any background knowledge of code, i want to implement this code, can anyone help me how to use this in IP Camera stream we have in my computer

采用yolov5烟雾-火灾2类检测模型，为自己的数据预标注，然后修正即可！

-OSError: libopencv_core.so.2.4: cannot open shared object file: No such file or directory 出现了这个错误，我不知道是为什么？有哪些原因吗？呜呜呜

您好，作者。我想咨询一个问题，Yolov5的烟火检测模型训练使用的数据规模是多少呢？

找了一下 没找到。

link：https://pan.baidu.com/s/1obRiPgI0-JJMJ-NFmFuJXA
code：52CV

tks: 52CV