automl
automl copied to clipboard
google
Google Brain AutoML
INT8 inference with TensorRT 8 with efficientdet-lite2 and efficientdet-lite1 detect nothing at all.
INT8 inference with TensorRT 8 with efficientdet-lite2 and efficientdet-lite1 in NVIDIA V100, Jetson Xavier NX and Jetson Xavier AGX detect nothing at all across all test images, no matter what...
det_advprop_tutorial.ipynb only have the eval code, i can't find the train script....
The guide for fine-tuning pascal from coco is changed and the `num_classes` is altered from 20 to 21 as below. ``` num_classes: 21 moving_average_decay: 0 label_id_mapping: {0: background, 1: aeroplane,...
@mingxingtan can we add mosaic aug in dataloader , as @fsx950223 suggested we can use dataset.batch(4).map(mosaic) im hoping this will increase some accuracy.
The D2 model when converted to int8 tflite I can leave it hours running to evaluate one single image without getting any output. The only output I get is the...
I have followed the instruction provided by @fsx950223 to create a int8 quantized tflite model. The quantization was for weights and layers output. The tflite obtained from a efficientdet-d2 checkpoint...
Is it possible to modify efficientDet lite architecture to output 4 points ( 8 values) instead of regular bboxes ? What do i have to change ? Any leads? Thanks
@fsx950223 @mingxingtan Could you explain this in more detail? Or Let me know any references? EfficientDets use a lot of GPU memory for a few reasons: Large input resolution: because...
``` Traceback (most recent call last): File "/home/fangsixie/.conda/envs/venv/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 2335, in load_weights hdf5_format.load_model_weights_from_hdf5_group(f, self, by_name, skip_mismatch) File "/home/fangsixie/.conda/envs/venv/lib/python3.6/site-packages/tensorflow/python/keras/saving/hdf5_format.py", line 733, in load_model_weights_from_hdf5_group f, model.layers, skip_mismatch=skip_mismatch) File "/home/fangsixie/.conda/envs/venv/lib/python3.6/site-packages/tensorflow/python/keras/saving/hdf5_format.py", line 822, in...
Hi, I have issues for running converted tflite int8 model from the available checkpoint. The following is the script: ```sh # download checkpoint in /data/efficientdet/pretrain wget https://storage.googleapis.com/cloud-tpu-checkpoints/efficientdet/coco/efficientdet-lite4.tgz tar xf efficientdet-lite4.tgz...
