add faster rcnn & vlp model

.gitattributes

@@ -29,3 +29,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 ./yolo/v5/yolov5m.bin filter=lfs diff=lfs merge=lfs -text
 ./yolo/v5/yolov5s.bin filter=lfs diff=lfs merge=lfs -text
 ./yolo/v5/yolov5x.bin filter=lfs diff=lfs merge=lfs -text
+./vlp/pytorch_model.bin filter=lfs diff=lfs merge=lfs -text
+./vlp/faster_rcnn_x101_64x4d_fpn_2x_model.bin filter=lfs diff=lfs merge=lfs -text
+./faster-rcnn/pascal-voc-detection/faster_rcnn_r50_c4_model.bin filter=lfs diff=lfs merge=lfs -text

faster-rcnn/pascal-voc-detection/faster_rcnn_r50_c4_config.yaml

@@ -0,0 +1,13 @@
+MODEL:
+  META_ARCHITECTURE: "GeneralizedRCNN"
+  RPN:
+    PRE_NMS_TOPK_TEST: 6000
+    POST_NMS_TOPK_TEST: 1000
+  ROI_HEADS:
+    NAME: "Res5ROIHeads"
+    NUM_CLASSES: 20
+  MASK_ON: False
+  RESNETS:
+    DEPTH: 50
+VERSION: 2
+

faster-rcnn/pascal-voc-detection/faster_rcnn_r50_c4_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6abec3dd4c28fc7a049add215e7877ba9bcf228f4f9e1150580b0a6abbe62790
+size 133221355

vlp/config.json

@@ -0,0 +1,18 @@
+{
+  "attention_probs_dropout_prob": 0.1,
+  "directionality": "bidi",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "max_position_embeddings": 512,
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "type_vocab_size": 6,
+  "layer_norm_epsilon": 0.00001,
+  "vocab_size": 28996,
+  "label_smoothing": 0.0,
+  "source_type_id": 4,
+  "target_type_id": 5
+}

vlp/faster_rcnn_x101_64x4d_fpn_2x_config.yaml

@@ -0,0 +1,42 @@
+MODEL:
+  META_ARCHITECTURE: "GeneralizedRCNN"
+  MASK_ON: False
+  BACKBONE:
+    NAME: "build_resnet_fpn_backbone"
+  RESNETS:
+    OUT_FEATURES: ["res2", "res3", "res4", "res5"]
+    STRIDE_IN_1X1: False  # this is a C2 model
+    NUM_GROUPS: 64
+    WIDTH_PER_GROUP: 4
+    DEPTH: 101
+  FPN:
+    IN_FEATURES: ["res2", "res3", "res4", "res5"]
+    OUT_CHANNELS: 512
+  ANCHOR_GENERATOR:
+    SIZES: [[32], [64], [128], [256], [512]]  # One size for each in feature map
+    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]  # Three aspect ratios (same for all in feature maps)
+  RPN:
+    IN_FEATURES: ["p2", "p3", "p4", "p5", "p6"]
+    PRE_NMS_TOPK_TRAIN: 2000  # Per FPN level
+    PRE_NMS_TOPK_TEST: 1000  # Per FPN level
+    # Detectron1 uses 2000 proposals per-batch,
+    # (See "modeling/rpn/rpn_outputs.py" for details of this legacy issue)
+    # which is approximately 1000 proposals per-image since the default batch size for FPN is 2.
+    POST_NMS_TOPK_TRAIN: 1000
+    POST_NMS_TOPK_TEST: 1000
+    NMS_THRESH: 0.7
+  ROI_HEADS:
+    NAME: "StandardROIHeads"
+    IN_FEATURES: ["p2", "p3", "p4", "p5"]
+    NUM_CLASSES: 1601
+    NMS_THRESH_TEST: 0.3
+  ROI_BOX_HEAD:
+    NAME: "FastRCNNConvFCHead"
+    NUM_CONV: 0
+    NUM_FC: 2
+    FC_DIM: 2048
+    POOLER_RESOLUTION: 7
+    POOLER_SAMPLING_RATIO: 2
+    POOLER_TYPE: "ROIAlign"
+VERSION: 2
+

vlp/faster_rcnn_x101_64x4d_fpn_2x_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ad57a9ca8cc157cc7d7bbe052602277ce4feed757b90efcaa050cdfee76d38d
+size 668770123

vlp/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e93d44ee113e6215d955646b4ed17539aad5deb5a898259d4120d23224520dd5
+size 463813740