Skip to content

Commit deb7af3

Browse files
rwightmanrwightman
committed
Merge branch 'master' into more_datasets
2 parents ce4e9bd + 2fc4e0e commit deb7af3

File tree

5 files changed

+296
-125
lines changed

5 files changed

+296
-125
lines changed

‎README.md

+98-112
Original file line numberDiff line numberDiff line change
@@ -17,6 +17,9 @@ Aside from the default model configs, there is a lot of flexibility to facilitat
1717

1818

1919
## Updates / Tasks
20+
### 2020-09-03
21+
* All models updated to latest checkpoints from TF original.
22+
* Add experimental soft-nms code, must be manually enabled right now. It is REALLY slow, .1-.2 mAP increase.
2023

2124
### 2020-07-27
2225
* Add updated TF ported weights for D3 model (better training) and model def and weights for new D7X model (54.3 val mAP)
@@ -160,27 +163,27 @@ If you are an organization is interested in sponsoring and any of this work, or
160163
| Variant | Download | mAP (val2017) | mAP (test-dev2017) | mAP (TF official val2017) | mAP (TF official test-dev2017) |
161164
| --- | --- | :---: | :---: | :---: | :---: |
162165
| lite0 | [tf_efficientdet_lite0.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_lite0-f5f303a9.pth) | 32.0 | TBD | N/A | N/A |
163-
| D0 | [tf_efficientdet_d0.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d0-d92fd44f.pth) | 33.6 | TBD | 33.5 | 33.8 |
164-
| D0 | [efficientdet_d0.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/efficientdet_d0-f3276ba8.pth) | 33.6 | TBD | 33.5 | 33.8 |
165-
| D1 | [tf_efficientdet_d1.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d1-4c7ebaf2.pth) | 39.3 | TBD | 39.1 | 39.6 |
166+
| D0 | [efficientdet_d0.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d0_34-f153e0cf.pth) | 33.6 | TBD | 33.5 | 33.8 |
167+
| D0 | [tf_efficientdet_d0.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d0_34-1851dfed.pth) | 34.2 | TBD | 34.3 | 34.6 |
166168
| D1 | [efficientdet_d1.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/efficientdet_d1-bb7e98fe.pth) | 39.4 | 39.5 | 39.1 | 39.6 |
167-
| D2 | [tf_efficientdet_d2.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d2-cb4ce77d.pth) | 42.6 | 43.1 | 42.5 | 43 |
169+
| D1 | [tf_efficientdet_d1.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d1_40-a30f94af.pth) | 40.1 | TBD | 40.2 | 40.5 |
170+
| D2 | [tf_efficientdet_d2.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d2_43-8107aa99.pth) | 43.4 | TBD | 42.5 | 43 |
168171
| D3 | [tf_efficientdet_d3.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d3_47-0b525f35.pth) | 47.1 | TBD | 47.2 | 47.5 |
169-
| D4 | [tf_efficientdet_d4.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d4-5b370b7a.pth) | 49.1 | TBD | 49.0 | 49.4 |
170-
| D5 | [tf_efficientdet_d5.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d5-ef44aea8.pth) | 50.4 | TBD | 50.5 | 50.7 |
171-
| D6 | [tf_efficientdet_d6.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d6-51cb0132.pth) | 51.2 | TBD | 51.3 | 51.7 |
172+
| D4 | [tf_efficientdet_d4.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d4_49-f56376d9.pth) | 49.2 | TBD | 49.3 | 49.7 |
173+
| D5 | [tf_efficientdet_d5.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d5_51-c79f9be6.pth) | 51.2 | TBD | 51.2 | 51.5 |
174+
| D6 | [tf_efficientdet_d6.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d6_52-4eda3773.pth) | 52.0 | TBD | 52.1 | 52.6 |
172175
| D7 | [tf_efficientdet_d7.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d7_53-6d1d7a95.pth) | 53.1 | 53.4 | 53.4 | 53.7 |
173176
| D7X | [tf_efficientdet_d7x.pth](https://github.com/rwightman/efficientdet-pytorch/releases/download/v0.1/tf_efficientdet_d7x-f390b87c.pth) | 54.3 | TBD | 54.4 | 55.1 |
174177

175-
_NOTE: Eval for TF D3, D7, and D7X numbers above were run with soft-nms, but still using normal NMS here._
178+
_NOTE: Official scores for all modules now using soft-nms, but still using normal NMS here._
176179

177180
## Usage
178181

179182
### Environment Setup
180183

181184
Tested in a Python 3.7 or 3.8 conda environment in Linux with:
182-
* PyTorch 1.4
183-
* PyTorch Image Models (timm) 0.1.20, `pip install timm` or local install from (https://github.com/rwightman/pytorch-image-models)
185+
* PyTorch 1.4 or PyTorch 1.6 (I recommend avoiding PyTorch 1.5 due to some jit and argmax issues)
186+
* PyTorch Image Models (timm) >= 0.1.28, `pip install timm` or local install from (https://github.com/rwightman/pytorch-image-models)
184187
* Apex AMP master (as of 2020-04)
185188

186189
*NOTE* - There is a conflict/bug with Numpy 1.18+ and pycocotools, force install numpy <= 1.17.5 or the coco eval will fail,
@@ -274,26 +277,9 @@ For this run I used some improved augmentations, still experimenting so not read
274277

275278
#### TEST-DEV2017
276279

277-
NOTE: I've only tried submitting D2 and D7 to dev server for sanity check so far
280+
NOTE: I've only tried submitting D7 to dev server for sanity check so far
278281

279-
##### EfficientDet-D2
280-
281-
```
282-
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.431
283-
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.624
284-
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.463
285-
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.226
286-
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.471
287-
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.585
288-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.345
289-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.543
290-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.575
291-
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.342
292-
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.632
293-
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.756
294-
```
295-
296-
##### EfficientDet-D7
282+
##### TF-EfficientDet-D7
297283
```
298284
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.534
299285
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.726
@@ -311,56 +297,56 @@ NOTE: I've only tried submitting D2 and D7 to dev server for sanity check so far
311297

312298
#### VAL2017
313299

314-
##### EfficientDet-D0
300+
##### TF-EfficientDet-D0
315301
```
316-
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.336
317-
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.516
318-
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.354
319-
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.125
320-
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.387
321-
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.528
322-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.288
323-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.440
324-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.467
325-
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.194
326-
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.549
327-
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.686
302+
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.341877
303+
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.525112
304+
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.360218
305+
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.131366
306+
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.399686
307+
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.537368
308+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.293137
309+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.447829
310+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.472954
311+
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.195282
312+
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.558127
313+
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.695312
328314
```
329315

330-
##### EfficientDet-D1
316+
##### TF-EfficientDet-D1
331317
```
332-
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.393
333-
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.583
334-
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.419
335-
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.187
336-
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.447
337-
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.572
338-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.323
339-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.501
340-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.532
341-
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.295
342-
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.599
343-
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.734
318+
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.401070
319+
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.590625
320+
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.422998
321+
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.211116
322+
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.459650
323+
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.577114
324+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.326565
325+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.507095
326+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.537278
327+
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.308963
328+
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.610450
329+
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.731814
344330
```
345331

346-
##### EfficientDet-D2
332+
##### TF-EfficientDet-D2
347333
```
348-
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.426
349-
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.618
350-
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.452
351-
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.237
352-
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.481
353-
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.590
354-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.342
355-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.537
356-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.569
357-
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.348
358-
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.633
359-
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.748
334+
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.434042
335+
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.627834
336+
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.463488
337+
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.237414
338+
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.486118
339+
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.606151
340+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.343016
341+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.538328
342+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.571489
343+
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.350301
344+
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.638884
345+
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.746671
360346
```
361347

362-
##### EfficientDet-D3
363-
_NOTE: Official TF impl uses soft-nms for their scoring of this model, not impl here yet_
348+
##### TF EfficientDet-D3
349+
364350
```
365351
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.471223
366352
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.661550
@@ -376,55 +362,55 @@ _NOTE: Official TF impl uses soft-nms for their scoring of this model, not impl
376362
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.779611
377363
```
378364

379-
##### EfficientDet-D4
365+
##### TF-EfficientDet-D4
380366
```
381-
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.491
382-
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.685
383-
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.531
384-
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.334
385-
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.539
386-
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.641
387-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.375
388-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.598
389-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.635
390-
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.468
391-
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.683
392-
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.780
367+
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.491759
368+
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.686005
369+
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.527791
370+
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.325658
371+
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.536508
372+
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.635309
373+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.373752
374+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.601733
375+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.638343
376+
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.463057
377+
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.685103
378+
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.789180
393379
```
394380

395-
##### EfficientDet-D5
381+
##### TF-EfficientDet-D5
396382
```
397-
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.504
398-
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.700
399-
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.543
400-
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.337
401-
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.549
402-
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.646
403-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.381
404-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.617
405-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.654
406-
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.485
407-
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.696
408-
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.791
383+
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.511767
384+
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.704835
385+
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.552920
386+
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.355680
387+
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.551341
388+
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.650184
389+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.384516
390+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.619196
391+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.657445
392+
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.499319
393+
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.695617
394+
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.788889
409395
```
410396

411-
##### EfficientDet-D6
397+
##### TF-EfficientDet-D6
412398
```
413-
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.512
414-
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.706
415-
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.551
416-
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.348
417-
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.555
418-
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.654
419-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.386
420-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.623
421-
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.661
422-
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.500
423-
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.701
424-
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.794
399+
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.520200
400+
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.713204
401+
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.560973
402+
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.361596
403+
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.567414
404+
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.657173
405+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.387733
406+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.629269
407+
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.667495
408+
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.499002
409+
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.711909
410+
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.802336
425411
```
426412

427-
##### EfficientDet-D7
413+
##### TF-EfficientDet-D7
428414
```
429415
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.531256
430416
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.724700
@@ -440,8 +426,8 @@ _NOTE: Official TF impl uses soft-nms for their scoring of this model, not impl
440426
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.806352
441427
```
442428

443-
##### EfficientDet-D7X
444-
_NOTE: Official TF impl uses soft-nms for their scoring of this model, not impl here yet_
429+
##### TF-EfficientDet-D7X
430+
445431
```
446432
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.543
447433
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.737

0 commit comments

Comments
 (0)