These are chat archives for beniz/deepdetect

15th
Feb 2017
sunatthegilddotcom
@sunatthegilddotcom
Feb 15 2017 05:41
@beniz , Now Tensorflow is built, I can setup service successfully by following your direction. curl -X PUT "http://localhost:8080/services/tensorincept" -d "{\"mllib\":\"tensorflow\",\"description\":\"image classification service\",\"type\":\"supervised\",\"parameters\":{\"input\":{\"connector\":\"image\",\"height\":224,\"width\":224},\"mllib\":{\"nclasses\":1001,\"inputlayer\":\"InputImage\"}},\"model\":{\"repository\":\"//home/wsun/deepdetect/models/tensorincept/\"}}"
{"status":{"code":201,"msg":"Created"}}
sunatthegilddotcom
@sunatthegilddotcom
Feb 15 2017 05:48
@beniz but prediction failed. curl -X POST "http://localhost:8080/predict" -d "{\"service\":\"tensorincept\",\"parameters\":{\"input\":{\"width\":224,\"height\":224},\"output\":{\"best\":3},\"mllib\":{\"gpu\":false}},\"data\":[\"https://deepdetect.com/img/cat.jpg\"]}"
{"status":{"code":500,"msg":"InternalError"}}

@beniz Dede console log is I /home/wsun/deepdetect/src/httpjsonapi.cc:363] Wed Feb 15 05:36:28 2017 UTC - 127.0.0.1 "PUT /services/tensorincept" 201 0

INFO - 05:35:21 - Ignoring source layer loss3/loss3
INFO - 05:38:11 - predict: using graphFile dir=//home/wsun/deepdetect/models/tensorincept//inception_resnet_v2.pb
INFO - 05:38:11 - using output layer=InceptionResnetV2/Logits/Predictions
W tensorflow/core/framework/op_kernel.cc:975] Invalid argument: computed output size would be negative
Invalid argument: computed output size would be negative
[Node: InceptionResnetV2/Logits/AvgPool_1a_8x8/AvgPool = AvgPool[T=DT_FLOAT, data_format="NHWC", ksize=[1, 8, 8, 1], padding="VALID", strides=[1, 1, 1, 1], _device="/job:localhost/replica:0/task:0/cpu:0"]]
E /home/wsun/deepdetect/src/services.h:490] service tensorincept mllib internal error: Invalid argument: computed output size would be negative
[Node: InceptionResnetV2/Logits/AvgPool_1a_8x8/AvgPool = AvgPool[T=DT_FLOAT, data_format="NHWC", ksize=[1, 8, 8, 1], padding="VALID", strides=[1, 1, 1, 1], _device="/job:localhost/replica:0/task:0/cpu:0"]]

E /home/wsun/deepdetect/src/httpjsonapi.cc:364] Wed Feb 15 05:38:12 2017 UTC - 127.0.0.1 "POST /predict" 500 2930

Emmanuel Benazera
@beniz
Feb 15 2017 06:34
never seen this but my guess is that you are using the wrong input size for this net, look at the tables we have in the readme.
Emmanuel Benazera
@beniz
Feb 15 2017 06:55
table is on the AMI documentation page, https://deepdetect.com/products/ami/
sunatthegilddotcom
@sunatthegilddotcom
Feb 15 2017 07:07
@beniz You are the best. I have to use 299x299 size. Thank you very much for your help.
Emmanuel Benazera
@beniz
Feb 15 2017 10:40
good