MobileNet-YOLO icon indicating copy to clipboard operation
MobileNet-YOLO copied to clipboard

How do I load images instead of LMDB?

Open GaryWBH opened this issue 4 years ago • 8 comments

HI eric612, I read your source code, there is a problem. Your Caffe reads LMDB type data. I need to directly read image data instead of LMDB. In the native Caffe, image reads do not support multiple tags and annotation boxes. How can I modify to read images with multiple tags using your model? Thanks

GaryWBH avatar Jun 09 '20 03:06 GaryWBH

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

WangYX-TKZ avatar Jun 09 '20 08:06 WangYX-TKZ

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

Wow,thanks, that's very kind of u!! I've been trying to modify Image_data_layer for days,but it doesn't work!! My Email: [email protected]

GaryWBH avatar Jun 09 '20 09:06 GaryWBH

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

Wow,thanks, that's very kind of u!! I've been trying to modify Image_data_layer for days,but it doesn't work!! My Email: [email protected]

I have send the Email to you. the file is suitable for my caffe project. May be your should modify somthing.

WangYX-TKZ avatar Jun 10 '20 13:06 WangYX-TKZ

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

Wow,thanks, that's very kind of u!! I've been trying to modify Image_data_layer for days,but it doesn't work!! My Email: [email protected]

I forgot the prototxt,I'm sure you should modify some details. example: layer { name: "input-data" type: "YoloV3RoiData" top: "data" top: "label" include { phase: TRAIN } window_data_param { source: "E:/Aera_data_en/test-Yolov3/data/transposed/self.trainval" root_folder: "E:/Aera_data_en/test-Yolov3/data/transposed/imageset/" cache_images: true } }

WangYX-TKZ avatar Jun 10 '20 13:06 WangYX-TKZ

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

Wow,thanks, that's very kind of u!! I've been trying to modify Image_data_layer for days,but it doesn't work!! My Email: [email protected]

I forgot the prototxt,I'm sure you should modify some details. example: layer { name: "input-data" type: "YoloV3RoiData" top: "data" top: "label" include { phase: TRAIN } window_data_param { source: "E:/Aera_data_en/test-Yolov3/data/transposed/self.trainval" root_folder: "E:/Aera_data_en/test-Yolov3/data/transposed/imageset/" cache_images: true } }

Got it~ I’m trying to read your code now, thank you for your help!

GaryWBH avatar Jun 10 '20 16:06 GaryWBH

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

Wow,thanks, that's very kind of u!! I've been trying to modify Image_data_layer for days,but it doesn't work!! My Email: [email protected]

I forgot the prototxt,I'm sure you should modify some details. example: layer { name: "input-data" type: "YoloV3RoiData" top: "data" top: "label" include { phase: TRAIN } window_data_param { source: "E:/Aera_data_en/test-Yolov3/data/transposed/self.trainval" root_folder: "E:/Aera_data_en/test-Yolov3/data/transposed/imageset/" cache_images: true } }

Got it~ I’m trying to read your code now, thank you for your help!

HI~WangYX-TKZ , It is very kind of u. and Can u share the image_data layer to me ? my email is [email protected]. I want this layer because I find that image_data_layer has better performance in some classification task. I guess the shuffle between epochs that lmdb hard to do is the reason.

applefishsky009 avatar Sep 22 '20 02:09 applefishsky009

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

Ths ! my email is [email protected]. I want this layer Because I have more than 3 channels images, I can finished base on your work. Ths!!!

superchenyan avatar Dec 21 '20 08:12 superchenyan

I have modified the image_data layer to support read images and annotation boxes. If you want the file, you could give me your Email.

Wow,thanks, that's very kind of u!! I've been trying to modify Image_data_layer for days,but it doesn't work!! My Email: [email protected]

HI~GaryWBH , Can u share the image_data layer to me ? my email is [email protected]. Ths !!!

superchenyan avatar Dec 26 '20 03:12 superchenyan